Warning: Permanently added '44.211.205.164' (ED25519) to the list of known hosts. You can reproduce this build on your computer by running: sudo dnf install copr-rpmbuild /usr/bin/copr-rpmbuild --verbose --drop-resultdir --task-url https://copr.fedorainfracloud.org/backend/get-build-task/11283205-fedora-43-x86_64 --chroot fedora-43-x86_64 Version: 1.11 PID: 11817 Logging PID: 11819 Task: {'allow_user_ssh': False, 'appstream': False, 'background': False, 'build_id': 11283205, 'buildroot_pkgs': [], 'chroot': 'fedora-43-x86_64', 'enable_net': False, 'fedora_review': False, 'git_hash': '0cbe8e54e1e505f67c52f3c806418e6d3d47e57c', 'git_repo': 'https://copr-dist-git.fedorainfracloud.org/git/mhough/neurofedora/python-array-api-extra', 'isolation': 'default', 'memory_reqs': 2048, 'package_name': 'python-array-api-extra', 'package_version': '0.11.5-1', 'project_dirname': 'neurofedora', 'project_name': 'neurofedora', 'project_owner': 'mhough', 'repo_priority': None, 'repos': [{'baseurl': 'https://download.copr.fedorainfracloud.org/results/mhough/neurofedora/fedora-43-x86_64/', 'id': 'copr_base', 'name': 'Copr repository', 'priority': None, 'repo_gpgcheck': False}], 'sandbox': 'mhough/neurofedora--mhough', 'source_json': {}, 'source_type': None, 'ssh_public_keys': None, 'storage': 1, 'submitter': 'mhough', 'tags': [], 'task_id': '11283205-fedora-43-x86_64', 'timeout': 86400, 'uses_devel_repo': False, 'with_opts': [], 'without_opts': []} Running: git clone https://copr-dist-git.fedorainfracloud.org/git/mhough/neurofedora/python-array-api-extra /var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra --depth 500 --no-single-branch --recursive cmd: ['git', 'clone', 'https://copr-dist-git.fedorainfracloud.org/git/mhough/neurofedora/python-array-api-extra', '/var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra', '--depth', '500', '--no-single-branch', '--recursive'] cwd: . rc: 0 stdout: stderr: Cloning into '/var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra'... Running: git checkout 0cbe8e54e1e505f67c52f3c806418e6d3d47e57c -- cmd: ['git', 'checkout', '0cbe8e54e1e505f67c52f3c806418e6d3d47e57c', '--'] cwd: /var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra rc: 0 stdout: stderr: Note: switching to '0cbe8e54e1e505f67c52f3c806418e6d3d47e57c'. You are in 'detached HEAD' state. You can look around, make experimental changes and commit them, and you can discard any commits you make in this state without impacting any branches by switching back to a branch. If you want to create a new branch to retain commits you create, you may do so (now or later) by using -c with the switch command. Example: git switch -c Or undo this operation with: git switch - Turn off this advice by setting config variable advice.detachedHead to false HEAD is now at 0cbe8e5 automatic import of python-array-api-extra Running: dist-git-client sources cmd: ['dist-git-client', 'sources'] cwd: /var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra rc: 0 stdout: stderr: INFO: Reading stdout from command: git rev-parse --abbrev-ref HEAD INFO: Reading stdout from command: git rev-parse HEAD INFO: Reading sources specification file: sources INFO: Downloading array_api_extra-0.11.5.tar.gz INFO: Reading stdout from command: curl --help all INFO: Calling: curl -H Pragma: -H 'Accept-Encoding: identity' -o array_api_extra-0.11.5.tar.gz --location --connect-timeout 60 --retry 3 --retry-delay 10 --remote-time --show-error --fail --retry-all-errors https://copr-dist-git.fedorainfracloud.org/repo/pkgs/mhough/neurofedora/python-array-api-extra/array_api_extra-0.11.5.tar.gz/md5/63668497670c299398f0e324b8901f3d/array_api_extra-0.11.5.tar.gz % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 110.1k 100 110.1k 0 0 10.09M 0 0 INFO: Reading stdout from command: md5sum array_api_extra-0.11.5.tar.gz tail: /var/lib/copr-rpmbuild/main.log: file truncated Running (timeout=86400): unbuffer mock --spec /var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra/python-array-api-extra.spec --sources /var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra --resultdir /var/lib/copr-rpmbuild/results --uniqueext 1794037731.754822 -r /var/lib/copr-rpmbuild/results/configs/child.cfg INFO: mock.py version 6.9 starting (python version = 3.14.5, NVR = mock-6.9-1.fc44), args: /usr/libexec/mock/mock --spec /var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra/python-array-api-extra.spec --sources /var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra --resultdir /var/lib/copr-rpmbuild/results --uniqueext 1794037731.754822 -r /var/lib/copr-rpmbuild/results/configs/child.cfg Start(bootstrap): init plugins INFO: tmpfs initialized INFO: selinux enabled INFO: chroot_scan: initialized INFO: compress_logs: initialized Finish(bootstrap): init plugins Start: init plugins INFO: tmpfs initialized INFO: selinux enabled INFO: chroot_scan: initialized INFO: compress_logs: initialized Finish: init plugins INFO: Signal handler active Start: run INFO: Start(/var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra/python-array-api-extra.spec) Config(fedora-43-x86_64) Start: clean chroot Finish: clean chroot Mock Version: 6.9 INFO: Mock Version: 6.9 Start(bootstrap): chroot init INFO: mounting tmpfs at /var/lib/mock/fedora-43-x86_64-bootstrap-1794037731.754822/root. INFO: calling preinit hooks INFO: enabled root cache INFO: enabled package manager cache Start(bootstrap): cleaning package manager metadata Finish(bootstrap): cleaning package manager metadata INFO: Guessed host environment type: unknown INFO: Using container image: registry.fedoraproject.org/fedora:43 INFO: Pulling image: registry.fedoraproject.org/fedora:43 INFO: Tagging container image as mock-bootstrap-473ee591-fdf3-47b9-8aa1-20a484067b36 INFO: Checking that 979205ea645d951faaa4407f56e249735a70898b436b16e9fed51b7faf113c55 image matches host's architecture INFO: Copy content of container 979205ea645d951faaa4407f56e249735a70898b436b16e9fed51b7faf113c55 to /var/lib/mock/fedora-43-x86_64-bootstrap-1794037731.754822/root INFO: mounting 979205ea645d951faaa4407f56e249735a70898b436b16e9fed51b7faf113c55 with podman image mount INFO: image 979205ea645d951faaa4407f56e249735a70898b436b16e9fed51b7faf113c55 as /var/lib/containers/storage/overlay/26542587b45449ba3598deb1df31d74cddb82eba4624a40a5485d08c3fb20845/merged INFO: umounting image 979205ea645d951faaa4407f56e249735a70898b436b16e9fed51b7faf113c55 (/var/lib/containers/storage/overlay/26542587b45449ba3598deb1df31d74cddb82eba4624a40a5485d08c3fb20845/merged) with podman image umount INFO: Removing image mock-bootstrap-473ee591-fdf3-47b9-8aa1-20a484067b36 INFO: Package manager dnf5 detected and used (fallback) INFO: Not updating bootstrap chroot, bootstrap_image_ready=True Start(bootstrap): creating root cache Finish(bootstrap): creating root cache Finish(bootstrap): chroot init Start: chroot init INFO: mounting tmpfs at /var/lib/mock/fedora-43-x86_64-1794037731.754822/root. INFO: calling preinit hooks INFO: enabled root cache INFO: enabled package manager cache Start: cleaning package manager metadata Finish: cleaning package manager metadata INFO: enabled HW Info plugin INFO: Package manager dnf5 detected and used (direct choice) INFO: Buildroot is handled by package management downloaded with a bootstrap image: rpm-6.0.2-1.fc43.x86_64 rpm-sequoia-1.10.3-1.fc43.x86_64 dnf5-5.2.18.0-5.fc43.x86_64 dnf5-plugins-5.2.18.0-5.fc43.x86_64 Start: installing minimal buildroot with dnf5 Updating and loading repositories: Copr repository 100% | 1.0 MiB/s | 672.4 KiB | 00m01s fedora 100% | 23.8 MiB/s | 35.4 MiB | 00m01s updates 100% | 5.7 MiB/s | 12.3 MiB | 00m02s Repositories loaded. Package Arch Version Repository Size Installing group/module packages: bash x86_64 5.3.0-2.fc43 fedora 8.4 MiB bzip2 x86_64 1.0.8-21.fc43 fedora 95.3 KiB coreutils x86_64 9.7-10.fc43 updates 5.4 MiB cpio x86_64 2.15-6.fc43 fedora 1.1 MiB diffutils x86_64 3.12-3.fc43 fedora 1.6 MiB fedora-release-common noarch 43-27 updates 20.6 KiB findutils x86_64 1:4.10.0-6.fc43 fedora 1.8 MiB gawk x86_64 5.3.2-2.fc43 fedora 1.8 MiB glibc-minimal-langpack x86_64 2.42-17.fc43 updates 0.0 B grep x86_64 3.12-2.fc43 fedora 1.0 MiB gzip x86_64 1.13-4.fc43 fedora 388.8 KiB info x86_64 7.2-7.fc43 updates 353.9 KiB patch x86_64 2.8-2.fc43 fedora 222.8 KiB redhat-rpm-config noarch 343-14.fc43 updates 182.9 KiB rpm-build x86_64 6.0.2-1.fc43 updates 286.4 KiB sed x86_64 4.9-5.fc43 fedora 857.3 KiB shadow-utils x86_64 2:4.18.0-3.fc43 fedora 3.9 MiB tar x86_64 2:1.35-6.fc43 fedora 2.9 MiB unzip x86_64 6.0-67.fc43 fedora 386.3 KiB util-linux x86_64 2.41.6-1.fc43 updates 3.5 MiB which x86_64 2.23-3.fc43 fedora 83.5 KiB xz x86_64 1:5.8.1-4.fc43 updates 1.3 MiB Installing dependencies: add-determinism x86_64 0.6.0-3.fc43 updates 2.3 MiB alternatives x86_64 1.33-3.fc43 updates 62.2 KiB ansible-srpm-macros noarch 1-18.1.fc43 fedora 35.7 KiB audit-libs x86_64 4.2.1-1.fc43 updates 382.5 KiB binutils x86_64 2.45.1-5.fc43 updates 27.3 MiB build-reproducibility-srpm-macros noarch 0.6.0-3.fc43 updates 735.0 B bzip2-libs x86_64 1.0.8-21.fc43 fedora 80.6 KiB ca-certificates noarch 2026.2.92_v9.0.319-2.fc43 updates 2.7 MiB cmake-srpm-macros noarch 3.31.11-1.fc43 updates 524.0 B coreutils-common x86_64 9.7-10.fc43 updates 11.3 MiB crypto-policies noarch 20251125-1.git63291f8.fc43 updates 147.1 KiB curl x86_64 8.15.0-12.fc43 updates 461.4 KiB cyrus-sasl-lib x86_64 2.1.28-33.fc43 fedora 2.3 MiB debugedit x86_64 5.2-3.fc43 fedora 214.0 KiB dwz x86_64 0.16-2.fc43 fedora 287.1 KiB ed x86_64 1.22.2-1.fc43 fedora 148.1 KiB efi-srpm-macros noarch 6-4.fc43 fedora 40.1 KiB elfutils x86_64 0.197-1.fc43 updates 3.1 MiB elfutils-debuginfod-client x86_64 0.197-1.fc43 updates 83.8 KiB elfutils-default-yama-scope noarch 0.197-1.fc43 updates 1.8 KiB elfutils-libelf x86_64 0.197-1.fc43 updates 1.6 MiB elfutils-libs x86_64 0.197-1.fc43 updates 695.4 KiB fedora-gpg-keys noarch 43-2 updates 259.0 KiB fedora-release noarch 43-27 updates 0.0 B fedora-release-identity-basic noarch 43-27 updates 631.0 B fedora-repos noarch 43-2 updates 4.9 KiB file x86_64 5.46-9.fc43 updates 100.2 KiB file-libs x86_64 5.46-9.fc43 updates 11.9 MiB filesystem x86_64 3.18-50.fc43 fedora 112.0 B filesystem-srpm-macros noarch 3.18-50.fc43 fedora 38.2 KiB fonts-srpm-macros noarch 1:2.0.5-23.fc43 fedora 55.8 KiB forge-srpm-macros noarch 0.4.0-3.fc43 fedora 38.9 KiB fpc-srpm-macros noarch 1.3-15.fc43 fedora 144.0 B gap-srpm-macros noarch 2-1.fc43 fedora 2.1 KiB gdb-minimal x86_64 18.1-1.fc43 updates 13.8 MiB gdbm-libs x86_64 1:1.23-10.fc43 fedora 129.9 KiB ghc-srpm-macros noarch 1.9.2-3.fc43 fedora 779.0 B glibc x86_64 2.42-17.fc43 updates 6.7 MiB glibc-common x86_64 2.42-17.fc43 updates 1.0 MiB glibc-gconv-extra x86_64 2.42-17.fc43 updates 7.2 MiB gmp x86_64 1:6.3.0-4.fc43 fedora 811.2 KiB gnat-srpm-macros noarch 7-1.fc43 updates 1.0 KiB gnome-srpm-macros noarch 1.0-1.fc43 updates 3.5 KiB gnulib-l10n noarch 20241231-1.fc43 fedora 655.0 KiB gnupg2 x86_64 2.4.9-5.fc43 updates 6.6 MiB gnupg2-dirmngr x86_64 2.4.9-5.fc43 updates 618.4 KiB gnupg2-gpg-agent x86_64 2.4.9-5.fc43 updates 671.4 KiB gnupg2-gpgconf x86_64 2.4.9-5.fc43 updates 250.0 KiB gnupg2-keyboxd x86_64 2.4.9-5.fc43 updates 201.4 KiB gnupg2-verify x86_64 2.4.9-5.fc43 updates 348.5 KiB gnutls x86_64 3.8.13-1.fc43 updates 3.8 MiB go-srpm-macros noarch 3.8.0-1.fc43 fedora 61.9 KiB gpgverify noarch 2.2-3.fc43 fedora 8.7 KiB ima-evm-utils-libs x86_64 1.6.2-6.fc43 fedora 60.7 KiB jansson x86_64 2.14-3.fc43 fedora 89.1 KiB java-srpm-macros noarch 1-7.fc43 fedora 870.0 B json-c x86_64 0.18-7.fc43 fedora 82.7 KiB kernel-srpm-macros noarch 1.0-27.fc43 fedora 1.9 KiB keyutils-libs x86_64 1.6.3-6.fc43 fedora 54.3 KiB krb5-libs x86_64 1.22.2-4.fc43 updates 2.3 MiB libacl x86_64 2.4.0-1.fc43 updates 39.9 KiB libarchive x86_64 3.8.4-1.fc43 updates 955.3 KiB libassuan x86_64 2.5.7-4.fc43 fedora 163.8 KiB libattr x86_64 2.6.0-1.fc43 updates 24.3 KiB libblkid x86_64 2.41.6-1.fc43 updates 262.3 KiB libbrotli x86_64 1.2.1-1.fc43 updates 865.0 KiB libcap x86_64 2.76-4.fc43 updates 209.1 KiB libcap-ng x86_64 0.9.6-1.fc43 updates 68.8 KiB libcom_err x86_64 1.47.3-2.fc43 fedora 63.1 KiB libcurl x86_64 8.15.0-12.fc43 updates 907.0 KiB libeconf x86_64 0.7.9-2.fc43 fedora 64.9 KiB libevent x86_64 2.1.13-1.fc43 updates 884.4 KiB libfdisk x86_64 2.41.6-1.fc43 updates 380.3 KiB libffi x86_64 3.5.2-1.fc43 updates 83.8 KiB libfsverity x86_64 1.6-3.fc43 fedora 28.5 KiB libgcc x86_64 15.3.1-1.fc43 updates 266.6 KiB libgcrypt x86_64 1.11.1-4.fc43 updates 1.6 MiB libgomp x86_64 15.3.1-1.fc43 updates 541.8 KiB libgpg-error x86_64 1.55-2.fc43 fedora 915.3 KiB libidn2 x86_64 2.3.8-2.fc43 fedora 552.5 KiB libksba x86_64 1.6.7-4.fc43 fedora 398.5 KiB liblastlog2 x86_64 2.41.6-1.fc43 updates 37.6 KiB libmount x86_64 2.41.6-1.fc43 updates 376.6 KiB libnghttp2 x86_64 1.66.0-3.fc43 updates 162.1 KiB libpfm x86_64 4.14.1-2.fc43 updates 2.4 MiB libpkgconf x86_64 2.3.0-3.fc43 fedora 78.1 KiB libpsl x86_64 0.21.5-6.fc43 fedora 76.4 KiB libselinux x86_64 3.9-5.fc43 fedora 193.1 KiB libsemanage x86_64 3.9-4.fc43 fedora 308.5 KiB libsepol x86_64 3.9-2.fc43 fedora 822.0 KiB libsmartcols x86_64 2.41.6-1.fc43 updates 180.3 KiB libssh x86_64 0.11.5-1.fc43 updates 573.1 KiB libssh-config noarch 0.11.5-1.fc43 updates 277.0 B libstdc++ x86_64 15.3.1-1.fc43 updates 2.8 MiB libtasn1 x86_64 4.21.0-1.fc43 updates 176.7 KiB libtool-ltdl x86_64 2.5.4-8.fc43 updates 70.1 KiB libunistring x86_64 1.1-10.fc43 fedora 1.7 MiB libusb1 x86_64 1.0.30-1.fc43 updates 176.0 KiB libuuid x86_64 2.41.6-1.fc43 updates 37.2 KiB libverto x86_64 0.3.2-11.fc43 fedora 25.4 KiB libxcrypt x86_64 4.5.2-1.fc43 updates 285.3 KiB libxml2 x86_64 2.12.10-5.fc43 fedora 1.7 MiB libzstd x86_64 1.5.7-2.fc43 fedora 799.9 KiB lua-libs x86_64 5.4.8-4.fc43 updates 281.9 KiB lua-srpm-macros noarch 1-16.fc43 fedora 1.3 KiB lz4-libs x86_64 1.10.0-3.fc43 fedora 161.4 KiB mpfr x86_64 4.2.2-2.fc43 fedora 832.8 KiB ncurses-base noarch 6.5-7.20250614.fc43 fedora 328.1 KiB ncurses-libs x86_64 6.5-7.20250614.fc43 fedora 946.3 KiB nettle x86_64 3.10.1-2.fc43 fedora 790.6 KiB npth x86_64 1.8-3.fc43 fedora 49.6 KiB ocaml-srpm-macros noarch 11-2.fc43 fedora 1.9 KiB openblas-srpm-macros noarch 2-20.fc43 fedora 112.0 B openldap x86_64 2.6.15-1.fc43 updates 659.2 KiB openssl-libs x86_64 1:3.5.9-1.fc43 updates 8.9 MiB p11-kit x86_64 0.26.5-1.fc43 updates 2.5 MiB p11-kit-trust x86_64 0.26.5-1.fc43 updates 466.3 KiB package-notes-srpm-macros noarch 0.20-1.fc43 updates 1.6 KiB pam-libs x86_64 1.7.1-5.fc43 updates 126.5 KiB pcre2 x86_64 10.47-1.fc43 updates 702.6 KiB pcre2-syntax noarch 10.47-1.fc43 updates 281.9 KiB perl-srpm-macros noarch 1-60.fc43 fedora 861.0 B pkgconf x86_64 2.3.0-3.fc43 fedora 88.5 KiB pkgconf-m4 noarch 2.3.0-3.fc43 fedora 14.4 KiB pkgconf-pkg-config x86_64 2.3.0-3.fc43 fedora 989.0 B popt x86_64 1.19-9.fc43 fedora 132.8 KiB publicsuffix-list-dafsa noarch 20260116-1.fc43 updates 70.4 KiB pyproject-srpm-macros noarch 1.24.0-1.fc43 updates 3.4 KiB python-srpm-macros noarch 3.14-5.fc43 fedora 51.5 KiB qt5-srpm-macros noarch 5.15.18-1.fc43 updates 500.0 B qt6-srpm-macros noarch 6.10.3-1.fc43 updates 472.0 B readline x86_64 8.3-2.fc43 fedora 511.7 KiB rpm x86_64 6.0.2-1.fc43 updates 3.1 MiB rpm-build-libs x86_64 6.0.2-1.fc43 updates 264.3 KiB rpm-libs x86_64 6.0.2-1.fc43 updates 933.5 KiB rpm-sequoia x86_64 1.10.3-1.fc43 updates 2.4 MiB rpm-sign-libs x86_64 6.0.2-1.fc43 updates 39.6 KiB rust-srpm-macros noarch 28.5-1.fc43 updates 5.5 KiB setup noarch 2.15.0-26.fc43 fedora 725.0 KiB sqlite-libs x86_64 3.50.2-3.fc43 updates 1.5 MiB systemd-libs x86_64 258.11-1.fc43 updates 2.3 MiB systemd-standalone-sysusers x86_64 258.11-1.fc43 updates 297.4 KiB tpm2-tss x86_64 4.1.3-8.fc43 fedora 1.6 MiB tree-sitter-srpm-macros noarch 0.4.2-1.fc43 fedora 8.3 KiB util-linux-core x86_64 2.41.6-1.fc43 updates 1.5 MiB xxhash-libs x86_64 0.8.4-2.fc43 updates 90.0 KiB xz-libs x86_64 1:5.8.1-4.fc43 updates 217.8 KiB zig-srpm-macros noarch 1-5.fc43 fedora 1.1 KiB zip x86_64 3.0-44.fc43 fedora 694.5 KiB zlib-ng-compat x86_64 2.3.3-2.fc43 updates 161.7 KiB zstd x86_64 1.5.7-2.fc43 fedora 1.7 MiB Installing groups: Buildsystem building group Transaction Summary: Installing: 173 packages Total size of inbound packages is 60 MiB. Need to download 60 MiB. After this operation, 204 MiB extra will be used (install 204 MiB, remove 0 B). [ 1/173] bzip2-0:1.0.8-21.fc43.x86_64 100% | 388.1 KiB/s | 51.6 KiB | 00m00s [ 2/173] cpio-0:2.15-6.fc43.x86_64 100% | 1.6 MiB/s | 293.1 KiB | 00m00s [ 3/173] diffutils-0:3.12-3.fc43.x86_6 100% | 3.9 MiB/s | 392.3 KiB | 00m00s [ 4/173] findutils-1:4.10.0-6.fc43.x86 100% | 8.8 MiB/s | 550.0 KiB | 00m00s [ 5/173] bash-0:5.3.0-2.fc43.x86_64 100% | 7.3 MiB/s | 1.9 MiB | 00m00s [ 6/173] gzip-0:1.13-4.fc43.x86_64 100% | 6.2 MiB/s | 170.1 KiB | 00m00s [ 7/173] patch-0:2.8-2.fc43.x86_64 100% | 4.6 MiB/s | 113.8 KiB | 00m00s [ 8/173] grep-0:3.12-2.fc43.x86_64 100% | 6.0 MiB/s | 299.1 KiB | 00m00s [ 9/173] sed-0:4.9-5.fc43.x86_64 100% | 10.0 MiB/s | 317.1 KiB | 00m00s [ 10/173] shadow-utils-2:4.18.0-3.fc43. 100% | 35.6 MiB/s | 1.3 MiB | 00m00s [ 11/173] unzip-0:6.0-67.fc43.x86_64 100% | 6.6 MiB/s | 183.7 KiB | 00m00s [ 12/173] tar-2:1.35-6.fc43.x86_64 100% | 14.7 MiB/s | 856.4 KiB | 00m00s [ 13/173] which-0:2.23-3.fc43.x86_64 100% | 1.7 MiB/s | 41.7 KiB | 00m00s [ 14/173] gawk-0:5.3.2-2.fc43.x86_64 100% | 32.1 MiB/s | 1.1 MiB | 00m00s [ 15/173] fedora-release-common-0:43-27 100% | 114.8 KiB/s | 22.6 KiB | 00m00s [ 16/173] glibc-minimal-langpack-0:2.42 100% | 212.2 KiB/s | 47.1 KiB | 00m00s [ 17/173] redhat-rpm-config-0:343-14.fc 100% | 768.9 KiB/s | 76.1 KiB | 00m00s [ 18/173] info-0:7.2-7.fc43.x86_64 100% | 1.0 MiB/s | 182.7 KiB | 00m00s [ 19/173] coreutils-0:9.7-10.fc43.x86_6 100% | 2.5 MiB/s | 1.1 MiB | 00m00s [ 20/173] rpm-build-0:6.0.2-1.fc43.x86_ 100% | 1.3 MiB/s | 135.3 KiB | 00m00s [ 21/173] filesystem-0:3.18-50.fc43.x86 100% | 36.0 MiB/s | 1.3 MiB | 00m00s [ 22/173] xz-1:5.8.1-4.fc43.x86_64 100% | 8.6 MiB/s | 572.7 KiB | 00m00s [ 23/173] ncurses-libs-0:6.5-7.20250614 100% | 12.0 MiB/s | 332.7 KiB | 00m00s [ 24/173] bzip2-libs-0:1.0.8-21.fc43.x8 100% | 1.8 MiB/s | 43.1 KiB | 00m00s [ 25/173] libselinux-0:3.9-5.fc43.x86_6 100% | 4.0 MiB/s | 97.7 KiB | 00m00s [ 26/173] ed-0:1.22.2-1.fc43.x86_64 100% | 3.4 MiB/s | 83.7 KiB | 00m00s [ 27/173] libeconf-0:0.7.9-2.fc43.x86_6 100% | 1.4 MiB/s | 35.2 KiB | 00m00s [ 28/173] util-linux-0:2.41.6-1.fc43.x8 100% | 5.7 MiB/s | 1.2 MiB | 00m00s [ 29/173] libsemanage-0:3.9-4.fc43.x86_ 100% | 4.6 MiB/s | 123.5 KiB | 00m00s [ 30/173] setup-0:2.15.0-26.fc43.noarch 100% | 6.1 MiB/s | 157.3 KiB | 00m00s [ 31/173] gmp-1:6.3.0-4.fc43.x86_64 100% | 10.1 MiB/s | 319.3 KiB | 00m00s [ 32/173] mpfr-0:4.2.2-2.fc43.x86_64 100% | 12.5 MiB/s | 347.0 KiB | 00m00s [ 33/173] readline-0:8.3-2.fc43.x86_64 100% | 8.4 MiB/s | 224.6 KiB | 00m00s [ 34/173] ansible-srpm-macros-0:1-18.1. 100% | 865.5 KiB/s | 19.9 KiB | 00m00s [ 35/173] glibc-common-0:2.42-17.fc43.x 100% | 5.8 MiB/s | 333.9 KiB | 00m00s [ 36/173] dwz-0:0.16-2.fc43.x86_64 100% | 5.5 MiB/s | 135.5 KiB | 00m00s [ 37/173] coreutils-common-0:9.7-10.fc4 100% | 17.9 MiB/s | 2.1 MiB | 00m00s [ 38/173] efi-srpm-macros-0:6-4.fc43.no 100% | 933.0 KiB/s | 22.4 KiB | 00m00s [ 39/173] filesystem-srpm-macros-0:3.18 100% | 1.1 MiB/s | 26.4 KiB | 00m00s [ 40/173] fonts-srpm-macros-1:2.0.5-23. 100% | 1.2 MiB/s | 27.2 KiB | 00m00s [ 41/173] forge-srpm-macros-0:0.4.0-3.f 100% | 837.0 KiB/s | 20.1 KiB | 00m00s [ 42/173] fpc-srpm-macros-0:1.3-15.fc43 100% | 343.1 KiB/s | 7.9 KiB | 00m00s [ 43/173] gap-srpm-macros-0:2-1.fc43.no 100% | 389.9 KiB/s | 9.0 KiB | 00m00s [ 44/173] ghc-srpm-macros-0:1.9.2-3.fc4 100% | 364.4 KiB/s | 8.7 KiB | 00m00s [ 45/173] go-srpm-macros-0:3.8.0-1.fc43 100% | 1.2 MiB/s | 28.3 KiB | 00m00s [ 46/173] java-srpm-macros-0:1-7.fc43.n 100% | 345.4 KiB/s | 7.9 KiB | 00m00s [ 47/173] kernel-srpm-macros-0:1.0-27.f 100% | 387.9 KiB/s | 8.9 KiB | 00m00s [ 48/173] lua-srpm-macros-0:1-16.fc43.n 100% | 380.7 KiB/s | 8.8 KiB | 00m00s [ 49/173] ocaml-srpm-macros-0:11-2.fc43 100% | 402.6 KiB/s | 9.3 KiB | 00m00s [ 50/173] openblas-srpm-macros-0:2-20.f 100% | 330.2 KiB/s | 7.6 KiB | 00m00s [ 51/173] perl-srpm-macros-0:1-60.fc43. 100% | 345.4 KiB/s | 8.3 KiB | 00m00s [ 52/173] python-srpm-macros-0:3.14-5.f 100% | 1.0 MiB/s | 23.4 KiB | 00m00s [ 53/173] tree-sitter-srpm-macros-0:0.4 100% | 556.3 KiB/s | 13.4 KiB | 00m00s [ 54/173] zig-srpm-macros-0:1-5.fc43.no 100% | 351.4 KiB/s | 8.4 KiB | 00m00s [ 55/173] zip-0:3.0-44.fc43.x86_64 100% | 8.8 MiB/s | 261.6 KiB | 00m00s [ 56/173] cmake-srpm-macros-0:3.31.11-1 100% | 209.7 KiB/s | 9.9 KiB | 00m00s [ 57/173] gnome-srpm-macros-0:1.0-1.fc4 100% | 183.2 KiB/s | 8.6 KiB | 00m00s [ 58/173] debugedit-0:5.2-3.fc43.x86_64 100% | 3.5 MiB/s | 85.6 KiB | 00m00s [ 59/173] popt-0:1.19-9.fc43.x86_64 100% | 2.6 MiB/s | 65.7 KiB | 00m00s [ 60/173] zstd-0:1.5.7-2.fc43.x86_64 100% | 13.2 MiB/s | 485.9 KiB | 00m00s [ 61/173] rpm-0:6.0.2-1.fc43.x86_64 100% | 10.4 MiB/s | 574.0 KiB | 00m00s [ 62/173] ncurses-base-0:6.5-7.20250614 100% | 3.4 MiB/s | 88.2 KiB | 00m00s [ 63/173] util-linux-core-0:2.41.6-1.fc 100% | 8.8 MiB/s | 551.9 KiB | 00m00s [ 64/173] xz-libs-1:5.8.1-4.fc43.x86_64 100% | 1.7 MiB/s | 112.8 KiB | 00m00s [ 65/173] libsepol-0:3.9-2.fc43.x86_64 100% | 12.5 MiB/s | 345.4 KiB | 00m00s [ 66/173] gnulib-l10n-0:20241231-1.fc43 100% | 5.6 MiB/s | 150.2 KiB | 00m00s [ 67/173] lz4-libs-0:1.10.0-3.fc43.x86_ 100% | 3.0 MiB/s | 78.0 KiB | 00m00s [ 68/173] rpm-libs-0:6.0.2-1.fc43.x86_6 100% | 5.7 MiB/s | 398.4 KiB | 00m00s [ 69/173] glibc-0:2.42-17.fc43.x86_64 100% | 29.1 MiB/s | 2.2 MiB | 00m00s [ 70/173] libzstd-0:1.5.7-2.fc43.x86_64 100% | 11.4 MiB/s | 314.6 KiB | 00m00s [ 71/173] rpm-build-libs-0:6.0.2-1.fc43 100% | 2.5 MiB/s | 124.1 KiB | 00m00s [ 72/173] rpm-sign-libs-0:6.0.2-1.fc43. 100% | 542.2 KiB/s | 25.5 KiB | 00m00s [ 73/173] ima-evm-utils-libs-0:1.6.2-6. 100% | 1.2 MiB/s | 29.3 KiB | 00m00s [ 74/173] libfsverity-0:1.6-3.fc43.x86_ 100% | 776.2 KiB/s | 18.6 KiB | 00m00s [ 75/173] tpm2-tss-0:4.1.3-8.fc43.x86_6 100% | 14.9 MiB/s | 425.9 KiB | 00m00s [ 76/173] json-c-0:0.18-7.fc43.x86_64 100% | 1.8 MiB/s | 45.0 KiB | 00m00s [ 77/173] glibc-gconv-extra-0:2.42-17.f 100% | 8.1 MiB/s | 1.6 MiB | 00m00s [ 78/173] libblkid-0:2.41.6-1.fc43.x86_ 100% | 2.5 MiB/s | 124.1 KiB | 00m00s [ 79/173] libuuid-0:2.41.6-1.fc43.x86_6 100% | 572.3 KiB/s | 26.9 KiB | 00m00s [ 80/173] libmount-0:2.41.6-1.fc43.x86_ 100% | 3.2 MiB/s | 164.5 KiB | 00m00s [ 81/173] libgcc-0:15.3.1-1.fc43.x86_64 100% | 2.9 MiB/s | 142.9 KiB | 00m00s [ 82/173] libsmartcols-0:2.41.6-1.fc43. 100% | 1.7 MiB/s | 84.6 KiB | 00m00s [ 83/173] libxcrypt-0:4.5.2-1.fc43.x86_ 100% | 2.6 MiB/s | 128.1 KiB | 00m00s [ 84/173] systemd-libs-0:258.11-1.fc43. 100% | 14.0 MiB/s | 832.0 KiB | 00m00s [ 85/173] audit-libs-0:4.2.1-1.fc43.x86 100% | 2.7 MiB/s | 139.6 KiB | 00m00s [ 86/173] libcap-ng-0:0.9.6-1.fc43.x86_ 100% | 729.8 KiB/s | 34.3 KiB | 00m00s [ 87/173] libfdisk-0:2.41.6-1.fc43.x86_ 100% | 3.3 MiB/s | 162.3 KiB | 00m00s [ 88/173] liblastlog2-0:2.41.6-1.fc43.x 100% | 516.5 KiB/s | 24.3 KiB | 00m00s [ 89/173] pam-libs-0:1.7.1-5.fc43.x86_6 100% | 1.2 MiB/s | 56.4 KiB | 00m00s [ 90/173] zlib-ng-compat-0:2.3.3-2.fc43 100% | 1.8 MiB/s | 88.7 KiB | 00m00s [ 91/173] libcap-0:2.76-4.fc43.x86_64 100% | 1.7 MiB/s | 86.1 KiB | 00m00s [ 92/173] gpgverify-0:2.2-3.fc43.noarch 100% | 462.6 KiB/s | 11.1 KiB | 00m00s [ 93/173] libassuan-0:2.5.7-4.fc43.x86_ 100% | 2.5 MiB/s | 67.4 KiB | 00m00s [ 94/173] libgpg-error-0:1.55-2.fc43.x8 100% | 8.8 MiB/s | 244.3 KiB | 00m00s [ 95/173] gnupg2-0:2.4.9-5.fc43.x86_64 100% | 20.5 MiB/s | 1.6 MiB | 00m00s [ 96/173] npth-0:1.8-3.fc43.x86_64 100% | 1.0 MiB/s | 25.7 KiB | 00m00s [ 97/173] gnupg2-dirmngr-0:2.4.9-5.fc43 100% | 5.4 MiB/s | 275.0 KiB | 00m00s [ 98/173] gnupg2-gpgconf-0:2.4.9-5.fc43 100% | 2.3 MiB/s | 115.3 KiB | 00m00s [ 99/173] gnupg2-gpg-agent-0:2.4.9-5.fc 100% | 4.9 MiB/s | 273.4 KiB | 00m00s [100/173] libksba-0:1.6.7-4.fc43.x86_64 100% | 6.0 MiB/s | 160.4 KiB | 00m00s [101/173] gnupg2-keyboxd-0:2.4.9-5.fc43 100% | 1.9 MiB/s | 94.7 KiB | 00m00s [102/173] gnupg2-verify-0:2.4.9-5.fc43. 100% | 3.4 MiB/s | 171.6 KiB | 00m00s [103/173] libacl-0:2.4.0-1.fc43.x86_64 100% | 554.8 KiB/s | 26.1 KiB | 00m00s [104/173] libstdc++-0:15.3.1-1.fc43.x86 100% | 12.8 MiB/s | 931.1 KiB | 00m00s [105/173] lua-libs-0:5.4.8-4.fc43.x86_6 100% | 2.6 MiB/s | 133.1 KiB | 00m00s [106/173] rpm-sequoia-0:1.10.3-1.fc43.x 100% | 15.6 MiB/s | 940.6 KiB | 00m00s [107/173] elfutils-libelf-0:0.197-1.fc4 100% | 4.3 MiB/s | 262.6 KiB | 00m00s [108/173] sqlite-libs-0:3.50.2-3.fc43.x 100% | 10.9 MiB/s | 758.9 KiB | 00m00s [109/173] elfutils-libs-0:0.197-1.fc43. 100% | 5.3 MiB/s | 273.0 KiB | 00m00s [110/173] elfutils-debuginfod-client-0: 100% | 986.1 KiB/s | 46.3 KiB | 00m00s [111/173] elfutils-0:0.197-1.fc43.x86_6 100% | 11.0 MiB/s | 655.7 KiB | 00m00s [112/173] file-libs-0:5.46-9.fc43.x86_6 100% | 14.3 MiB/s | 849.9 KiB | 00m00s [113/173] file-0:5.46-9.fc43.x86_64 100% | 1.0 MiB/s | 48.5 KiB | 00m00s [114/173] libgomp-0:15.3.1-1.fc43.x86_6 100% | 7.1 MiB/s | 383.6 KiB | 00m00s [115/173] jansson-0:2.14-3.fc43.x86_64 100% | 1.8 MiB/s | 45.3 KiB | 00m00s [116/173] libxml2-0:2.12.10-5.fc43.x86_ 100% | 21.8 MiB/s | 692.7 KiB | 00m00s [117/173] libarchive-0:3.8.4-1.fc43.x86 100% | 6.6 MiB/s | 422.6 KiB | 00m00s [118/173] pkgconf-pkg-config-0:2.3.0-3. 100% | 417.8 KiB/s | 9.6 KiB | 00m00s [119/173] pkgconf-0:2.3.0-3.fc43.x86_64 100% | 1.8 MiB/s | 44.6 KiB | 00m00s [120/173] pkgconf-m4-0:2.3.0-3.fc43.noa 100% | 604.8 KiB/s | 13.9 KiB | 00m00s [121/173] binutils-0:2.45.1-5.fc43.x86_ 100% | 49.1 MiB/s | 5.9 MiB | 00m00s [122/173] libpkgconf-0:2.3.0-3.fc43.x86 100% | 1.5 MiB/s | 37.9 KiB | 00m00s [123/173] curl-0:8.15.0-12.fc43.x86_64 100% | 4.2 MiB/s | 226.8 KiB | 00m00s [124/173] build-reproducibility-srpm-ma 100% | 253.8 KiB/s | 11.9 KiB | 00m00s [125/173] add-determinism-0:0.6.0-3.fc4 100% | 14.3 MiB/s | 893.7 KiB | 00m00s [126/173] gnat-srpm-macros-0:7-1.fc43.n 100% | 184.2 KiB/s | 8.7 KiB | 00m00s [127/173] package-notes-srpm-macros-0:0 100% | 214.2 KiB/s | 9.9 KiB | 00m00s [128/173] pyproject-srpm-macros-0:1.24. 100% | 327.7 KiB/s | 15.4 KiB | 00m00s [129/173] qt5-srpm-macros-0:5.15.18-1.f 100% | 187.0 KiB/s | 8.6 KiB | 00m00s [130/173] qt6-srpm-macros-0:6.10.3-1.fc 100% | 194.4 KiB/s | 9.1 KiB | 00m00s [131/173] rust-srpm-macros-0:28.5-1.fc4 100% | 238.2 KiB/s | 11.2 KiB | 00m00s [132/173] libgcrypt-0:1.11.1-4.fc43.x86 100% | 9.7 MiB/s | 596.0 KiB | 00m00s [133/173] libidn2-0:2.3.8-2.fc43.x86_64 100% | 6.6 MiB/s | 174.9 KiB | 00m00s [134/173] gnutls-0:3.8.13-1.fc43.x86_64 100% | 19.6 MiB/s | 1.4 MiB | 00m00s [135/173] openssl-libs-1:3.5.9-1.fc43.x 100% | 22.1 MiB/s | 2.6 MiB | 00m00s [136/173] libunistring-0:1.1-10.fc43.x8 100% | 18.3 MiB/s | 542.9 KiB | 00m00s [137/173] nettle-0:3.10.1-2.fc43.x86_64 100% | 14.3 MiB/s | 424.2 KiB | 00m00s [138/173] gdbm-libs-1:1.23-10.fc43.x86_ 100% | 2.1 MiB/s | 56.8 KiB | 00m00s [139/173] cyrus-sasl-lib-0:2.1.28-33.fc 100% | 24.8 MiB/s | 787.9 KiB | 00m00s [140/173] openldap-0:2.6.15-1.fc43.x86_ 100% | 5.1 MiB/s | 259.7 KiB | 00m00s [141/173] fedora-repos-0:43-2.noarch 100% | 188.9 KiB/s | 8.9 KiB | 00m00s [142/173] fedora-gpg-keys-0:43-2.noarch 100% | 3.7 MiB/s | 184.2 KiB | 00m00s [143/173] elfutils-default-yama-scope-0 100% | 244.1 KiB/s | 11.5 KiB | 00m00s [144/173] libpfm-0:4.14.1-2.fc43.x86_64 100% | 5.4 MiB/s | 275.3 KiB | 00m00s [145/173] libattr-0:2.6.0-1.fc43.x86_64 100% | 388.1 KiB/s | 17.9 KiB | 00m00s [146/173] alternatives-0:1.33-3.fc43.x8 100% | 864.3 KiB/s | 40.6 KiB | 00m00s [147/173] libusb1-0:1.0.30-1.fc43.x86_6 100% | 1.7 MiB/s | 80.5 KiB | 00m00s [148/173] pcre2-0:10.47-1.fc43.x86_64 100% | 5.1 MiB/s | 267.2 KiB | 00m00s [149/173] pcre2-syntax-0:10.47-1.fc43.n 100% | 3.4 MiB/s | 164.7 KiB | 00m00s [150/173] crypto-policies-0:20251125-1. 100% | 1.8 MiB/s | 97.8 KiB | 00m00s [151/173] ca-certificates-0:2026.2.92_v 100% | 16.2 MiB/s | 980.2 KiB | 00m00s [152/173] libevent-0:2.1.13-1.fc43.x86_ 100% | 5.1 MiB/s | 259.5 KiB | 00m00s [153/173] libtool-ltdl-0:2.5.4-8.fc43.x 100% | 771.2 KiB/s | 36.2 KiB | 00m00s [154/173] libtasn1-0:4.21.0-1.fc43.x86_ 100% | 1.5 MiB/s | 74.5 KiB | 00m00s [155/173] p11-kit-0:0.26.5-1.fc43.x86_6 100% | 9.8 MiB/s | 542.8 KiB | 00m00s [156/173] p11-kit-trust-0:0.26.5-1.fc43 100% | 2.8 MiB/s | 140.6 KiB | 00m00s [157/173] libffi-0:3.5.2-1.fc43.x86_64 100% | 855.4 KiB/s | 41.1 KiB | 00m00s [158/173] systemd-standalone-sysusers-0 100% | 3.0 MiB/s | 149.4 KiB | 00m00s [159/173] xxhash-libs-0:0.8.4-2.fc43.x8 100% | 514.2 KiB/s | 38.6 KiB | 00m00s [160/173] fedora-release-0:43-27.noarch 100% | 189.3 KiB/s | 11.5 KiB | 00m00s [161/173] gdb-minimal-0:18.1-1.fc43.x86 100% | 46.5 MiB/s | 4.5 MiB | 00m00s [162/173] fedora-release-identity-basic 100% | 262.0 KiB/s | 12.3 KiB | 00m00s [163/173] libpsl-0:0.21.5-6.fc43.x86_64 100% | 2.2 MiB/s | 65.0 KiB | 00m00s [164/173] libcurl-0:8.15.0-12.fc43.x86_ 100% | 7.7 MiB/s | 403.9 KiB | 00m00s [165/173] libssh-0:0.11.5-1.fc43.x86_64 100% | 4.7 MiB/s | 234.6 KiB | 00m00s [166/173] libssh-config-0:0.11.5-1.fc43 100% | 190.3 KiB/s | 8.9 KiB | 00m00s [167/173] publicsuffix-list-dafsa-0:202 100% | 1.3 MiB/s | 60.3 KiB | 00m00s [168/173] keyutils-libs-0:1.6.3-6.fc43. 100% | 1.3 MiB/s | 31.4 KiB | 00m00s [169/173] libcom_err-0:1.47.3-2.fc43.x8 100% | 1.1 MiB/s | 26.8 KiB | 00m00s [170/173] libverto-0:0.3.2-11.fc43.x86_ 100% | 861.5 KiB/s | 20.7 KiB | 00m00s [171/173] krb5-libs-0:1.22.2-4.fc43.x86 100% | 13.4 MiB/s | 770.4 KiB | 00m00s [172/173] libbrotli-0:1.2.1-1.fc43.x86_ 100% | 6.8 MiB/s | 349.5 KiB | 00m00s [173/173] libnghttp2-0:1.66.0-3.fc43.x8 100% | 1.5 MiB/s | 71.0 KiB | 00m00s -------------------------------------------------------------------------------- [173/173] Total 100% | 19.2 MiB/s | 59.9 MiB | 00m03s Running transaction Importing OpenPGP key 0x31645531: UserID : "Fedora (43) " Fingerprint: C6E7F081CF80E13146676E88829B606631645531 From : file:///usr/share/distribution-gpg-keys/fedora/RPM-GPG-KEY-fedora-43-primary The key was successfully imported. [ 1/175] Verify package files 100% | 819.0 B/s | 173.0 B | 00m00s [ 2/175] Prepare transaction 100% | 4.6 KiB/s | 173.0 B | 00m00s [ 3/175] Installing libgcc-0:15.3.1-1. 100% | 262.0 MiB/s | 268.3 KiB | 00m00s [ 4/175] Installing publicsuffix-list- 100% | 0.0 B/s | 71.1 KiB | 00m00s [ 5/175] Installing libssh-config-0:0. 100% | 0.0 B/s | 816.0 B | 00m00s [ 6/175] Installing fedora-release-ide 100% | 867.2 KiB/s | 888.0 B | 00m00s [ 7/175] Installing fedora-gpg-keys-0: 100% | 57.8 MiB/s | 355.0 KiB | 00m00s [ 8/175] Installing fedora-repos-0:43- 100% | 0.0 B/s | 5.7 KiB | 00m00s [ 9/175] Installing fedora-release-com 100% | 24.3 MiB/s | 24.8 KiB | 00m00s [ 10/175] Installing fedora-release-0:4 100% | 15.1 KiB/s | 124.0 B | 00m00s >>> Running sysusers scriptlet: setup-0:2.15.0-26.fc43.noarch >>> Finished sysusers scriptlet: setup-0:2.15.0-26.fc43.noarch >>> Scriptlet output: >>> Creating group 'adm' with GID 4. >>> Creating group 'audio' with GID 63. >>> Creating group 'cdrom' with GID 11. >>> Creating group 'clock' with GID 103. >>> Creating group 'dialout' with GID 18. >>> Creating group 'disk' with GID 6. >>> Creating group 'floppy' with GID 19. >>> Creating group 'ftp' with GID 50. >>> Creating group 'games' with GID 20. >>> Creating group 'input' with GID 104. >>> Creating group 'kmem' with GID 9. >>> Creating group 'kvm' with GID 36. >>> Creating group 'lock' with GID 54. >>> Creating group 'lp' with GID 7. >>> Creating group 'mail' with GID 12. >>> Creating group 'man' with GID 15. >>> Creating group 'mem' with GID 8. >>> Creating group 'nobody' with GID 65534. >>> Creating group 'render' with GID 105. >>> Creating group 'root' with GID 0. >>> Creating group 'sgx' with GID 106. >>> Creating group 'sys' with GID 3. >>> Creating group 'tape' with GID 33. >>> Creating group 'tty' with GID 5. >>> Creating group 'users' with GID 100. >>> Creating group 'utmp' with GID 22. >>> Creating group 'video' with GID 39. >>> Creating group 'wheel' with GID 10. >>> Creating user 'adm' (adm) with UID 3 and GID 4. >>> Creating group 'bin' with GID 1. >>> Creating user 'bin' (bin) with UID 1 and GID 1. >>> Creating group 'daemon' with GID 2. >>> Creating user 'daemon' (daemon) with UID 2 and GID 2. >>> Creating user 'ftp' (FTP User) with UID 14 and GID 50. >>> Creating user 'games' (games) with UID 12 and GID 100. >>> Creating user 'halt' (halt) with UID 7 and GID 0. >>> Creating user 'lp' (lp) with UID 4 and GID 7. >>> Creating user 'mail' (mail) with UID 8 and GID 12. >>> Creating user 'nobody' (Kernel Overflow User) with UID 65534 and GID 65534. >>> Creating user 'operator' (operator) with UID 11 and GID 0. >>> Creating user 'root' (Super User) with UID 0 and GID 0. >>> Creating user 'shutdown' (shutdown) with UID 6 and GID 0. >>> Creating user 'sync' (sync) with UID 5 and GID 0. >>> [ 11/175] Installing setup-0:2.15.0-26. 100% | 54.9 MiB/s | 730.6 KiB | 00m00s >>> [RPM] /etc/hosts created as /etc/hosts.rpmnew [ 12/175] Installing filesystem-0:3.18- 100% | 3.3 MiB/s | 212.8 KiB | 00m00s [ 13/175] Installing pcre2-syntax-0:10. 100% | 277.7 MiB/s | 284.3 KiB | 00m00s [ 14/175] Installing rust-srpm-macros-0 100% | 0.0 B/s | 6.4 KiB | 00m00s [ 15/175] Installing qt6-srpm-macros-0: 100% | 0.0 B/s | 748.0 B | 00m00s [ 16/175] Installing qt5-srpm-macros-0: 100% | 0.0 B/s | 776.0 B | 00m00s [ 17/175] Installing package-notes-srpm 100% | 0.0 B/s | 2.1 KiB | 00m00s [ 18/175] Installing gnat-srpm-macros-0 100% | 0.0 B/s | 1.3 KiB | 00m00s [ 19/175] Installing pkgconf-m4-0:2.3.0 100% | 0.0 B/s | 14.8 KiB | 00m00s [ 20/175] Installing gnulib-l10n-0:2024 100% | 215.5 MiB/s | 661.9 KiB | 00m00s [ 21/175] Installing coreutils-common-0 100% | 451.7 MiB/s | 11.3 MiB | 00m00s [ 22/175] Installing ncurses-base-0:6.5 100% | 115.1 MiB/s | 353.5 KiB | 00m00s [ 23/175] Installing bash-0:5.3.0-2.fc4 100% | 301.1 MiB/s | 8.4 MiB | 00m00s [ 24/175] Installing glibc-common-0:2.4 100% | 67.9 MiB/s | 1.0 MiB | 00m00s [ 25/175] Installing glibc-gconv-extra- 100% | 330.9 MiB/s | 7.3 MiB | 00m00s [ 26/175] Installing glibc-0:2.42-17.fc 100% | 202.8 MiB/s | 6.7 MiB | 00m00s [ 27/175] Installing ncurses-libs-0:6.5 100% | 310.1 MiB/s | 952.8 KiB | 00m00s [ 28/175] Installing glibc-minimal-lang 100% | 0.0 B/s | 124.0 B | 00m00s [ 29/175] Installing zlib-ng-compat-0:2 100% | 158.7 MiB/s | 162.6 KiB | 00m00s [ 30/175] Installing bzip2-libs-0:1.0.8 100% | 0.0 B/s | 81.7 KiB | 00m00s [ 31/175] Installing libgpg-error-0:1.5 100% | 64.3 MiB/s | 921.1 KiB | 00m00s [ 32/175] Installing libstdc++-0:15.3.1 100% | 407.0 MiB/s | 2.8 MiB | 00m00s [ 33/175] Installing xz-libs-1:5.8.1-4. 100% | 213.8 MiB/s | 218.9 KiB | 00m00s [ 34/175] Installing libassuan-0:2.5.7- 100% | 161.7 MiB/s | 165.6 KiB | 00m00s [ 35/175] Installing libgcrypt-0:1.11.1 100% | 393.7 MiB/s | 1.6 MiB | 00m00s [ 36/175] Installing readline-0:8.3-2.f 100% | 501.8 MiB/s | 513.9 KiB | 00m00s [ 37/175] Installing gmp-1:6.3.0-4.fc43 100% | 397.2 MiB/s | 813.5 KiB | 00m00s [ 38/175] Installing libuuid-0:2.41.6-1 100% | 0.0 B/s | 38.4 KiB | 00m00s [ 39/175] Installing libacl-0:2.4.0-1.f 100% | 0.0 B/s | 40.7 KiB | 00m00s [ 40/175] Installing popt-0:1.19-9.fc43 100% | 68.1 MiB/s | 139.4 KiB | 00m00s [ 41/175] Installing libzstd-0:1.5.7-2. 100% | 391.2 MiB/s | 801.1 KiB | 00m00s [ 42/175] Installing elfutils-libelf-0: 100% | 542.8 MiB/s | 1.6 MiB | 00m00s [ 43/175] Installing npth-0:1.8-3.fc43. 100% | 0.0 B/s | 50.7 KiB | 00m00s [ 44/175] Installing libblkid-0:2.41.6- 100% | 257.3 MiB/s | 263.5 KiB | 00m00s [ 45/175] Installing libxcrypt-0:4.5.2- 100% | 281.3 MiB/s | 288.0 KiB | 00m00s [ 46/175] Installing sqlite-libs-0:3.50 100% | 379.1 MiB/s | 1.5 MiB | 00m00s [ 47/175] Installing gnupg2-gpgconf-0:2 100% | 20.5 MiB/s | 252.1 KiB | 00m00s [ 48/175] Installing json-c-0:0.18-7.fc 100% | 0.0 B/s | 84.0 KiB | 00m00s [ 49/175] Installing libunistring-0:1.1 100% | 431.7 MiB/s | 1.7 MiB | 00m00s [ 50/175] Installing libidn2-0:2.3.8-2. 100% | 272.8 MiB/s | 558.7 KiB | 00m00s [ 51/175] Installing libtasn1-0:4.21.0- 100% | 21.8 MiB/s | 178.6 KiB | 00m00s [ 52/175] Installing crypto-policies-0: 100% | 42.0 MiB/s | 172.2 KiB | 00m00s [ 53/175] Installing gnupg2-verify-0:2. 100% | 28.5 MiB/s | 349.9 KiB | 00m00s [ 54/175] Installing dwz-0:0.16-2.fc43. 100% | 25.6 MiB/s | 288.5 KiB | 00m00s [ 55/175] Installing mpfr-0:4.2.2-2.fc4 100% | 407.4 MiB/s | 834.4 KiB | 00m00s [ 56/175] Installing gawk-0:5.3.2-2.fc4 100% | 113.5 MiB/s | 1.8 MiB | 00m00s [ 57/175] Installing libksba-0:1.6.7-4. 100% | 391.7 MiB/s | 401.1 KiB | 00m00s [ 58/175] Installing unzip-0:6.0-67.fc4 100% | 34.6 MiB/s | 389.8 KiB | 00m00s [ 59/175] Installing file-libs-0:5.46-9 100% | 741.3 MiB/s | 11.9 MiB | 00m00s [ 60/175] Installing file-0:5.46-9.fc43 100% | 9.0 MiB/s | 101.7 KiB | 00m00s [ 61/175] Installing libeconf-0:0.7.9-2 100% | 0.0 B/s | 66.5 KiB | 00m00s [ 62/175] Installing libsepol-0:3.9-2.f 100% | 401.8 MiB/s | 822.9 KiB | 00m00s [ 63/175] Installing lz4-libs-0:1.10.0- 100% | 158.6 MiB/s | 162.5 KiB | 00m00s [ 64/175] Installing libsmartcols-0:2.4 100% | 177.2 MiB/s | 181.5 KiB | 00m00s [ 65/175] Installing libcap-ng-0:0.9.6- 100% | 69.0 MiB/s | 70.6 KiB | 00m00s [ 66/175] Installing audit-libs-0:4.2.1 100% | 375.9 MiB/s | 384.9 KiB | 00m00s [ 67/175] Installing pam-libs-0:1.7.1-5 100% | 125.7 MiB/s | 128.8 KiB | 00m00s [ 68/175] Installing libcap-0:2.76-4.fc 100% | 19.0 MiB/s | 214.3 KiB | 00m00s [ 69/175] Installing systemd-libs-0:258 100% | 391.3 MiB/s | 2.3 MiB | 00m00s [ 70/175] Installing lua-libs-0:5.4.8-4 100% | 276.7 MiB/s | 283.3 KiB | 00m00s [ 71/175] Installing libattr-0:2.6.0-1. 100% | 0.0 B/s | 25.2 KiB | 00m00s [ 72/175] Installing alternatives-0:1.3 100% | 5.7 MiB/s | 63.8 KiB | 00m00s [ 73/175] Installing pcre2-0:10.47-1.fc 100% | 343.8 MiB/s | 704.1 KiB | 00m00s [ 74/175] Installing libselinux-0:3.9-5 100% | 189.8 MiB/s | 194.4 KiB | 00m00s [ 75/175] Installing grep-0:3.12-2.fc43 100% | 71.6 MiB/s | 1.0 MiB | 00m00s [ 76/175] Installing findutils-1:4.10.0 100% | 123.9 MiB/s | 1.9 MiB | 00m00s [ 77/175] Installing sed-0:4.9-5.fc43.x 100% | 60.4 MiB/s | 865.5 KiB | 00m00s [ 78/175] Installing xz-1:5.8.1-4.fc43. 100% | 88.8 MiB/s | 1.3 MiB | 00m00s [ 79/175] Installing libmount-0:2.41.6- 100% | 368.9 MiB/s | 377.8 KiB | 00m00s [ 80/175] Installing libffi-0:3.5.2-1.f 100% | 83.2 MiB/s | 85.2 KiB | 00m00s [ 81/175] Installing p11-kit-0:0.26.5-1 100% | 149.7 MiB/s | 2.5 MiB | 00m00s [ 82/175] Installing p11-kit-trust-0:0. 100% | 24.0 MiB/s | 467.9 KiB | 00m00s [ 83/175] Installing openssl-libs-1:3.5 100% | 425.8 MiB/s | 8.9 MiB | 00m00s [ 84/175] Installing coreutils-0:9.7-10 100% | 181.0 MiB/s | 5.4 MiB | 00m00s [ 85/175] Installing ca-certificates-0: 100% | 2.5 MiB/s | 2.5 MiB | 00m01s [ 86/175] Installing gzip-0:1.13-4.fc43 100% | 29.6 MiB/s | 394.4 KiB | 00m00s [ 87/175] Installing libfsverity-0:1.6- 100% | 0.0 B/s | 29.5 KiB | 00m00s [ 88/175] Installing rpm-sequoia-0:1.10 100% | 408.5 MiB/s | 2.5 MiB | 00m00s [ 89/175] Installing libevent-0:2.1.13- 100% | 289.1 MiB/s | 888.2 KiB | 00m00s [ 90/175] Installing util-linux-core-0: 100% | 92.2 MiB/s | 1.5 MiB | 00m00s [ 91/175] Installing tar-2:1.35-6.fc43. 100% | 164.3 MiB/s | 3.0 MiB | 00m00s [ 92/175] Installing libsemanage-0:3.9- 100% | 303.0 MiB/s | 310.2 KiB | 00m00s [ 93/175] Installing systemd-standalone 100% | 26.5 MiB/s | 298.1 KiB | 00m00s [ 94/175] Installing rpm-libs-0:6.0.2-1 100% | 304.4 MiB/s | 935.1 KiB | 00m00s [ 95/175] Installing libusb1-0:1.0.30-1 100% | 21.7 MiB/s | 177.7 KiB | 00m00s >>> Running sysusers scriptlet: tpm2-tss-0:4.1.3-8.fc43.x86_64 >>> Finished sysusers scriptlet: tpm2-tss-0:4.1.3-8.fc43.x86_64 >>> Scriptlet output: >>> Creating group 'tss' with GID 59. >>> Creating user 'tss' (Account used for TPM access) with UID 59 and GID 59. >>> [ 96/175] Installing tpm2-tss-0:4.1.3-8 100% | 314.4 MiB/s | 1.6 MiB | 00m00s [ 97/175] Installing ima-evm-utils-libs 100% | 60.5 MiB/s | 62.0 KiB | 00m00s [ 98/175] Installing gnupg2-gpg-agent-0 100% | 33.0 MiB/s | 675.4 KiB | 00m00s [ 99/175] Installing liblastlog2-0:2.41 100% | 6.5 MiB/s | 39.8 KiB | 00m00s [100/175] Installing zstd-0:1.5.7-2.fc4 100% | 114.0 MiB/s | 1.7 MiB | 00m00s [101/175] Installing zip-0:3.0-44.fc43. 100% | 56.8 MiB/s | 698.4 KiB | 00m00s [102/175] Installing gnupg2-keyboxd-0:2 100% | 33.0 MiB/s | 202.7 KiB | 00m00s [103/175] Installing libpsl-0:0.21.5-6. 100% | 75.7 MiB/s | 77.5 KiB | 00m00s [104/175] Installing libfdisk-0:2.41.6- 100% | 372.5 MiB/s | 381.4 KiB | 00m00s [105/175] Installing nettle-0:3.10.1-2. 100% | 387.5 MiB/s | 793.7 KiB | 00m00s [106/175] Installing gnutls-0:3.8.13-1. 100% | 418.6 MiB/s | 3.8 MiB | 00m00s [107/175] Installing libxml2-0:2.12.10- 100% | 106.5 MiB/s | 1.7 MiB | 00m00s [108/175] Installing libarchive-0:3.8.4 100% | 311.6 MiB/s | 957.2 KiB | 00m00s [109/175] Installing bzip2-0:1.0.8-21.f 100% | 8.9 MiB/s | 99.8 KiB | 00m00s [110/175] Installing add-determinism-0: 100% | 143.9 MiB/s | 2.3 MiB | 00m00s [111/175] Installing build-reproducibil 100% | 0.0 B/s | 1.0 KiB | 00m00s [112/175] Installing cpio-0:2.15-6.fc43 100% | 78.5 MiB/s | 1.1 MiB | 00m00s [113/175] Installing diffutils-0:3.12-3 100% | 104.1 MiB/s | 1.6 MiB | 00m00s [114/175] Installing ed-0:1.22.2-1.fc43 100% | 13.3 MiB/s | 150.4 KiB | 00m00s [115/175] Installing patch-0:2.8-2.fc43 100% | 19.9 MiB/s | 224.3 KiB | 00m00s [116/175] Installing libgomp-0:15.3.1-1 100% | 530.4 MiB/s | 543.2 KiB | 00m00s [117/175] Installing jansson-0:2.14-3.f 100% | 88.3 MiB/s | 90.5 KiB | 00m00s [118/175] Installing libpkgconf-0:2.3.0 100% | 0.0 B/s | 79.2 KiB | 00m00s [119/175] Installing pkgconf-0:2.3.0-3. 100% | 8.1 MiB/s | 91.0 KiB | 00m00s [120/175] Installing pkgconf-pkg-config 100% | 161.2 KiB/s | 1.8 KiB | 00m00s [121/175] Installing gdbm-libs-1:1.23-1 100% | 128.5 MiB/s | 131.6 KiB | 00m00s [122/175] Installing cyrus-sasl-lib-0:2 100% | 143.5 MiB/s | 2.3 MiB | 00m00s [123/175] Installing libpfm-0:4.14.1-2. 100% | 592.1 MiB/s | 2.4 MiB | 00m00s [124/175] Installing libtool-ltdl-0:2.5 100% | 69.6 MiB/s | 71.2 KiB | 00m00s [125/175] Installing openldap-0:2.6.15- 100% | 323.7 MiB/s | 663.0 KiB | 00m00s [126/175] Installing gnupg2-dirmngr-0:2 100% | 31.9 MiB/s | 621.1 KiB | 00m00s [127/175] Installing gnupg2-0:2.4.9-5.f 100% | 252.5 MiB/s | 6.6 MiB | 00m00s [128/175] Installing rpm-sign-libs-0:6. 100% | 0.0 B/s | 40.5 KiB | 00m00s [129/175] Installing gpgverify-0:2.2-3. 100% | 0.0 B/s | 9.4 KiB | 00m00s [130/175] Installing xxhash-libs-0:0.8. 100% | 89.2 MiB/s | 91.4 KiB | 00m00s [131/175] Installing keyutils-libs-0:1. 100% | 0.0 B/s | 55.7 KiB | 00m00s [132/175] Installing libcom_err-0:1.47. 100% | 0.0 B/s | 64.2 KiB | 00m00s [133/175] Installing libverto-0:0.3.2-1 100% | 26.6 MiB/s | 27.2 KiB | 00m00s [134/175] Installing krb5-libs-0:1.22.2 100% | 388.7 MiB/s | 2.3 MiB | 00m00s [135/175] Installing libssh-0:0.11.5-1. 100% | 280.8 MiB/s | 575.1 KiB | 00m00s [136/175] Installing libbrotli-0:1.2.1- 100% | 423.5 MiB/s | 867.3 KiB | 00m00s [137/175] Installing libnghttp2-0:1.66. 100% | 159.4 MiB/s | 163.2 KiB | 00m00s [138/175] Installing libcurl-0:8.15.0-1 100% | 295.6 MiB/s | 908.1 KiB | 00m00s [139/175] Installing curl-0:8.15.0-12.f 100% | 21.6 MiB/s | 463.9 KiB | 00m00s [140/175] Installing rpm-0:6.0.2-1.fc43 100% | 92.0 MiB/s | 2.6 MiB | 00m00s [141/175] Installing efi-srpm-macros-0: 100% | 40.2 MiB/s | 41.1 KiB | 00m00s [142/175] Installing java-srpm-macros-0 100% | 0.0 B/s | 1.1 KiB | 00m00s [143/175] Installing lua-srpm-macros-0: 100% | 0.0 B/s | 1.9 KiB | 00m00s [144/175] Installing tree-sitter-srpm-m 100% | 0.0 B/s | 9.3 KiB | 00m00s [145/175] Installing zig-srpm-macros-0: 100% | 0.0 B/s | 1.7 KiB | 00m00s [146/175] Installing cmake-srpm-macros- 100% | 0.0 B/s | 804.0 B | 00m00s [147/175] Installing filesystem-srpm-ma 100% | 0.0 B/s | 38.9 KiB | 00m00s [148/175] Installing elfutils-default-y 100% | 408.6 KiB/s | 2.0 KiB | 00m00s [149/175] Installing elfutils-libs-0:0. 100% | 227.0 MiB/s | 697.2 KiB | 00m00s [150/175] Installing elfutils-debuginfo 100% | 7.0 MiB/s | 86.0 KiB | 00m00s [151/175] Installing elfutils-0:0.197-1 100% | 173.5 MiB/s | 3.1 MiB | 00m00s [152/175] Installing binutils-0:2.45.1- 100% | 365.0 MiB/s | 27.4 MiB | 00m00s [153/175] Installing gdb-minimal-0:18.1 100% | 321.5 MiB/s | 13.8 MiB | 00m00s [154/175] Installing debugedit-0:5.2-3. 100% | 17.7 MiB/s | 217.3 KiB | 00m00s [155/175] Installing rpm-build-libs-0:6 100% | 258.9 MiB/s | 265.1 KiB | 00m00s [156/175] Installing gnome-srpm-macros- 100% | 0.0 B/s | 4.1 KiB | 00m00s [157/175] Installing perl-srpm-macros-0 100% | 0.0 B/s | 1.1 KiB | 00m00s [158/175] Installing openblas-srpm-macr 100% | 0.0 B/s | 392.0 B | 00m00s [159/175] Installing ocaml-srpm-macros- 100% | 0.0 B/s | 2.1 KiB | 00m00s [160/175] Installing kernel-srpm-macros 100% | 0.0 B/s | 2.3 KiB | 00m00s [161/175] Installing ghc-srpm-macros-0: 100% | 0.0 B/s | 1.0 KiB | 00m00s [162/175] Installing gap-srpm-macros-0: 100% | 0.0 B/s | 2.7 KiB | 00m00s [163/175] Installing fpc-srpm-macros-0: 100% | 0.0 B/s | 420.0 B | 00m00s [164/175] Installing ansible-srpm-macro 100% | 0.0 B/s | 36.2 KiB | 00m00s [165/175] Installing rpm-build-0:6.0.2- 100% | 22.2 MiB/s | 295.6 KiB | 00m00s [166/175] Installing pyproject-srpm-mac 100% | 0.0 B/s | 4.0 KiB | 00m00s [167/175] Installing redhat-rpm-config- 100% | 184.7 MiB/s | 189.1 KiB | 00m00s [168/175] Installing forge-srpm-macros- 100% | 0.0 B/s | 40.3 KiB | 00m00s [169/175] Installing fonts-srpm-macros- 100% | 0.0 B/s | 57.0 KiB | 00m00s [170/175] Installing go-srpm-macros-0:3 100% | 0.0 B/s | 63.0 KiB | 00m00s [171/175] Installing python-srpm-macros 100% | 51.6 MiB/s | 52.8 KiB | 00m00s [172/175] Installing util-linux-0:2.41. 100% | 108.0 MiB/s | 3.6 MiB | 00m00s [173/175] Installing shadow-utils-2:4.1 100% | 158.8 MiB/s | 4.0 MiB | 00m00s [174/175] Installing which-0:2.23-3.fc4 100% | 7.6 MiB/s | 85.7 KiB | 00m00s [175/175] Installing info-0:7.2-7.fc43. 100% | 275.1 KiB/s | 354.3 KiB | 00m01s Complete! Finish: installing minimal buildroot with dnf5 Start: creating root cache Finish: creating root cache Finish: chroot init INFO: Installed packages: INFO: add-determinism-0.6.0-3.fc43.x86_64 alternatives-1.33-3.fc43.x86_64 ansible-srpm-macros-1-18.1.fc43.noarch audit-libs-4.2.1-1.fc43.x86_64 bash-5.3.0-2.fc43.x86_64 binutils-2.45.1-5.fc43.x86_64 build-reproducibility-srpm-macros-0.6.0-3.fc43.noarch bzip2-1.0.8-21.fc43.x86_64 bzip2-libs-1.0.8-21.fc43.x86_64 ca-certificates-2026.2.92_v9.0.319-2.fc43.noarch cmake-srpm-macros-3.31.11-1.fc43.noarch coreutils-9.7-10.fc43.x86_64 coreutils-common-9.7-10.fc43.x86_64 cpio-2.15-6.fc43.x86_64 crypto-policies-20251125-1.git63291f8.fc43.noarch curl-8.15.0-12.fc43.x86_64 cyrus-sasl-lib-2.1.28-33.fc43.x86_64 debugedit-5.2-3.fc43.x86_64 diffutils-3.12-3.fc43.x86_64 dwz-0.16-2.fc43.x86_64 ed-1.22.2-1.fc43.x86_64 efi-srpm-macros-6-4.fc43.noarch elfutils-0.197-1.fc43.x86_64 elfutils-debuginfod-client-0.197-1.fc43.x86_64 elfutils-default-yama-scope-0.197-1.fc43.noarch elfutils-libelf-0.197-1.fc43.x86_64 elfutils-libs-0.197-1.fc43.x86_64 fedora-gpg-keys-43-2.noarch fedora-release-43-27.noarch fedora-release-common-43-27.noarch fedora-release-identity-basic-43-27.noarch fedora-repos-43-2.noarch file-5.46-9.fc43.x86_64 file-libs-5.46-9.fc43.x86_64 filesystem-3.18-50.fc43.x86_64 filesystem-srpm-macros-3.18-50.fc43.noarch findutils-4.10.0-6.fc43.x86_64 fonts-srpm-macros-2.0.5-23.fc43.noarch forge-srpm-macros-0.4.0-3.fc43.noarch fpc-srpm-macros-1.3-15.fc43.noarch gap-srpm-macros-2-1.fc43.noarch gawk-5.3.2-2.fc43.x86_64 gdb-minimal-18.1-1.fc43.x86_64 gdbm-libs-1.23-10.fc43.x86_64 ghc-srpm-macros-1.9.2-3.fc43.noarch glibc-2.42-17.fc43.x86_64 glibc-common-2.42-17.fc43.x86_64 glibc-gconv-extra-2.42-17.fc43.x86_64 glibc-minimal-langpack-2.42-17.fc43.x86_64 gmp-6.3.0-4.fc43.x86_64 gnat-srpm-macros-7-1.fc43.noarch gnome-srpm-macros-1.0-1.fc43.noarch gnulib-l10n-20241231-1.fc43.noarch gnupg2-2.4.9-5.fc43.x86_64 gnupg2-dirmngr-2.4.9-5.fc43.x86_64 gnupg2-gpg-agent-2.4.9-5.fc43.x86_64 gnupg2-gpgconf-2.4.9-5.fc43.x86_64 gnupg2-keyboxd-2.4.9-5.fc43.x86_64 gnupg2-verify-2.4.9-5.fc43.x86_64 gnutls-3.8.13-1.fc43.x86_64 go-srpm-macros-3.8.0-1.fc43.noarch gpg-pubkey-c6e7f081cf80e13146676e88829b606631645531-66b6dccf gpgverify-2.2-3.fc43.noarch grep-3.12-2.fc43.x86_64 gzip-1.13-4.fc43.x86_64 ima-evm-utils-libs-1.6.2-6.fc43.x86_64 info-7.2-7.fc43.x86_64 jansson-2.14-3.fc43.x86_64 java-srpm-macros-1-7.fc43.noarch json-c-0.18-7.fc43.x86_64 kernel-srpm-macros-1.0-27.fc43.noarch keyutils-libs-1.6.3-6.fc43.x86_64 krb5-libs-1.22.2-4.fc43.x86_64 libacl-2.4.0-1.fc43.x86_64 libarchive-3.8.4-1.fc43.x86_64 libassuan-2.5.7-4.fc43.x86_64 libattr-2.6.0-1.fc43.x86_64 libblkid-2.41.6-1.fc43.x86_64 libbrotli-1.2.1-1.fc43.x86_64 libcap-2.76-4.fc43.x86_64 libcap-ng-0.9.6-1.fc43.x86_64 libcom_err-1.47.3-2.fc43.x86_64 libcurl-8.15.0-12.fc43.x86_64 libeconf-0.7.9-2.fc43.x86_64 libevent-2.1.13-1.fc43.x86_64 libfdisk-2.41.6-1.fc43.x86_64 libffi-3.5.2-1.fc43.x86_64 libfsverity-1.6-3.fc43.x86_64 libgcc-15.3.1-1.fc43.x86_64 libgcrypt-1.11.1-4.fc43.x86_64 libgomp-15.3.1-1.fc43.x86_64 libgpg-error-1.55-2.fc43.x86_64 libidn2-2.3.8-2.fc43.x86_64 libksba-1.6.7-4.fc43.x86_64 liblastlog2-2.41.6-1.fc43.x86_64 libmount-2.41.6-1.fc43.x86_64 libnghttp2-1.66.0-3.fc43.x86_64 libpfm-4.14.1-2.fc43.x86_64 libpkgconf-2.3.0-3.fc43.x86_64 libpsl-0.21.5-6.fc43.x86_64 libselinux-3.9-5.fc43.x86_64 libsemanage-3.9-4.fc43.x86_64 libsepol-3.9-2.fc43.x86_64 libsmartcols-2.41.6-1.fc43.x86_64 libssh-0.11.5-1.fc43.x86_64 libssh-config-0.11.5-1.fc43.noarch libstdc++-15.3.1-1.fc43.x86_64 libtasn1-4.21.0-1.fc43.x86_64 libtool-ltdl-2.5.4-8.fc43.x86_64 libunistring-1.1-10.fc43.x86_64 libusb1-1.0.30-1.fc43.x86_64 libuuid-2.41.6-1.fc43.x86_64 libverto-0.3.2-11.fc43.x86_64 libxcrypt-4.5.2-1.fc43.x86_64 libxml2-2.12.10-5.fc43.x86_64 libzstd-1.5.7-2.fc43.x86_64 lua-libs-5.4.8-4.fc43.x86_64 lua-srpm-macros-1-16.fc43.noarch lz4-libs-1.10.0-3.fc43.x86_64 mpfr-4.2.2-2.fc43.x86_64 ncurses-base-6.5-7.20250614.fc43.noarch ncurses-libs-6.5-7.20250614.fc43.x86_64 nettle-3.10.1-2.fc43.x86_64 npth-1.8-3.fc43.x86_64 ocaml-srpm-macros-11-2.fc43.noarch openblas-srpm-macros-2-20.fc43.noarch openldap-2.6.15-1.fc43.x86_64 openssl-libs-3.5.9-1.fc43.x86_64 p11-kit-0.26.5-1.fc43.x86_64 p11-kit-trust-0.26.5-1.fc43.x86_64 package-notes-srpm-macros-0.20-1.fc43.noarch pam-libs-1.7.1-5.fc43.x86_64 patch-2.8-2.fc43.x86_64 pcre2-10.47-1.fc43.x86_64 pcre2-syntax-10.47-1.fc43.noarch perl-srpm-macros-1-60.fc43.noarch pkgconf-2.3.0-3.fc43.x86_64 pkgconf-m4-2.3.0-3.fc43.noarch pkgconf-pkg-config-2.3.0-3.fc43.x86_64 popt-1.19-9.fc43.x86_64 publicsuffix-list-dafsa-20260116-1.fc43.noarch pyproject-srpm-macros-1.24.0-1.fc43.noarch python-srpm-macros-3.14-5.fc43.noarch qt5-srpm-macros-5.15.18-1.fc43.noarch qt6-srpm-macros-6.10.3-1.fc43.noarch readline-8.3-2.fc43.x86_64 redhat-rpm-config-343-14.fc43.noarch rpm-6.0.2-1.fc43.x86_64 rpm-build-6.0.2-1.fc43.x86_64 rpm-build-libs-6.0.2-1.fc43.x86_64 rpm-libs-6.0.2-1.fc43.x86_64 rpm-sequoia-1.10.3-1.fc43.x86_64 rpm-sign-libs-6.0.2-1.fc43.x86_64 rust-srpm-macros-28.5-1.fc43.noarch sed-4.9-5.fc43.x86_64 setup-2.15.0-26.fc43.noarch shadow-utils-4.18.0-3.fc43.x86_64 sqlite-libs-3.50.2-3.fc43.x86_64 systemd-libs-258.11-1.fc43.x86_64 systemd-standalone-sysusers-258.11-1.fc43.x86_64 tar-1.35-6.fc43.x86_64 tpm2-tss-4.1.3-8.fc43.x86_64 tree-sitter-srpm-macros-0.4.2-1.fc43.noarch unzip-6.0-67.fc43.x86_64 util-linux-2.41.6-1.fc43.x86_64 util-linux-core-2.41.6-1.fc43.x86_64 which-2.23-3.fc43.x86_64 xxhash-libs-0.8.4-2.fc43.x86_64 xz-5.8.1-4.fc43.x86_64 xz-libs-5.8.1-4.fc43.x86_64 zig-srpm-macros-1-5.fc43.noarch zip-3.0-44.fc43.x86_64 zlib-ng-compat-2.3.3-2.fc43.x86_64 zstd-1.5.7-2.fc43.x86_64 Start: buildsrpm Start: rpmbuild -bs Building target platforms: x86_64 Building for target x86_64 setting SOURCE_DATE_EPOCH=1793923200 Wrote: /builddir/build/SRPMS/python-array-api-extra-0.11.5-1.fc43.src.rpm Finish: rpmbuild -bs INFO: chroot_scan: 1 files copied to /var/lib/copr-rpmbuild/results/chroot_scan INFO: /var/lib/mock/fedora-43-x86_64-1794037731.754822/root/var/log/dnf5.log INFO: chroot_scan: creating tarball /var/lib/copr-rpmbuild/results/chroot_scan.tar.gz /bin/tar: Removing leading `/' from member names Finish: buildsrpm INFO: Done(/var/lib/copr-rpmbuild/workspace/workdir-5nak1ttb/python-array-api-extra/python-array-api-extra.spec) Config(child) 0 minutes 20 seconds INFO: Results and/or logs in: /var/lib/copr-rpmbuild/results INFO: Cleaning up build root ('cleanup_on_success=True') Start: clean chroot INFO: unmounting tmpfs. Finish: clean chroot INFO: Start(/var/lib/copr-rpmbuild/results/python-array-api-extra-0.11.5-1.fc43.src.rpm) Config(fedora-43-x86_64) Start(bootstrap): chroot init INFO: mounting tmpfs at /var/lib/mock/fedora-43-x86_64-bootstrap-1794037731.754822/root. INFO: reusing tmpfs at /var/lib/mock/fedora-43-x86_64-bootstrap-1794037731.754822/root. INFO: calling preinit hooks INFO: enabled root cache INFO: enabled package manager cache Start(bootstrap): cleaning package manager metadata Finish(bootstrap): cleaning package manager metadata Finish(bootstrap): chroot init Start: chroot init INFO: mounting tmpfs at /var/lib/mock/fedora-43-x86_64-1794037731.754822/root. INFO: calling preinit hooks INFO: enabled root cache Start: unpacking root cache Finish: unpacking root cache INFO: enabled package manager cache Start: cleaning package manager metadata Finish: cleaning package manager metadata INFO: enabled HW Info plugin INFO: Buildroot is handled by package management downloaded with a bootstrap image: rpm-6.0.2-1.fc43.x86_64 rpm-sequoia-1.10.3-1.fc43.x86_64 dnf5-5.2.18.0-5.fc43.x86_64 dnf5-plugins-5.2.18.0-5.fc43.x86_64 Finish: chroot init Start: build phase for python-array-api-extra-0.11.5-1.fc43.src.rpm Start: build setup for python-array-api-extra-0.11.5-1.fc43.src.rpm Building target platforms: x86_64 Building for target x86_64 setting SOURCE_DATE_EPOCH=1793923200 Wrote: /builddir/build/SRPMS/python-array-api-extra-0.11.5-1.fc43.src.rpm Updating and loading repositories: Copr repository 100% | 10.3 KiB/s | 1.5 KiB | 00m00s fedora 100% | 57.4 KiB/s | 23.5 KiB | 00m00s updates 100% | 65.8 KiB/s | 18.6 KiB | 00m00s Repositories loaded. Package Arch Version Repository Size Installing: pyproject-rpm-macros noarch 1.24.0-1.fc43 updates 169.0 KiB python3-array-api-strict noarch 2.0.1-7.fc43 fedora 694.1 KiB python3-devel x86_64 3.14.8-1.fc43 updates 2.0 MiB python3-hypothesis noarch 6.123.0-7.fc43 fedora 5.0 MiB python3-numpy x86_64 1:2.3.5-1.fc43 updates 41.4 MiB python3-pytest noarch 8.3.5-9.fc43 fedora 21.6 MiB python3-typing-extensions noarch 4.16.0-4.fc43 updates 552.0 KiB Installing dependencies: expat x86_64 2.9.0-1.fc43 updates 337.6 KiB flexiblas x86_64 3.5.0-1.fc43 updates 38.0 KiB flexiblas-netlib x86_64 3.5.0-1.fc43 updates 16.5 MiB flexiblas-openblas-openmp x86_64 3.5.0-1.fc43 updates 39.2 KiB libgfortran x86_64 15.3.1-1.fc43 updates 3.3 MiB libquadmath x86_64 15.3.1-1.fc43 updates 317.9 KiB mpdecimal x86_64 4.0.1-2.fc43 fedora 217.2 KiB openblas x86_64 0.3.29-2.fc43 fedora 111.7 KiB openblas-openmp x86_64 0.3.29-2.fc43 fedora 43.7 MiB python-pip-wheel noarch 25.1.1-19.fc43 updates 1.2 MiB python-rpm-macros noarch 3.14-5.fc43 fedora 23.2 KiB python3 x86_64 3.14.8-1.fc43 updates 28.7 KiB python3-attrs noarch 25.4.0-1.fc43 updates 436.8 KiB python3-iniconfig noarch 2.1.0-3.fc43 fedora 42.4 KiB python3-libs x86_64 3.14.8-1.fc43 updates 43.5 MiB python3-numpy-f2py x86_64 1:2.3.5-1.fc43 updates 2.1 MiB python3-packaging noarch 25.0-7.fc43 fedora 607.4 KiB python3-pluggy noarch 1.6.0-4.fc43 fedora 211.0 KiB python3-rpm-generators noarch 14-13.fc43 fedora 81.7 KiB python3-rpm-macros noarch 3.14-5.fc43 fedora 6.5 KiB python3-sortedcontainers noarch 2.4.0-25.fc43 fedora 392.9 KiB tzdata noarch 2026e-1.fc43 updates 1.6 MiB Transaction Summary: Installing: 29 packages Total size of inbound packages is 37 MiB. Need to download 37 MiB. After this operation, 186 MiB extra will be used (install 186 MiB, remove 0 B). [ 1/29] python3-array-api-strict-0:2.0. 100% | 950.3 KiB/s | 164.4 KiB | 00m00s [ 2/29] python3-pytest-0:8.3.5-9.fc43.n 100% | 8.1 MiB/s | 2.2 MiB | 00m00s [ 3/29] python3-hypothesis-0:6.123.0-7. 100% | 4.4 MiB/s | 1.2 MiB | 00m00s [ 4/29] pyproject-rpm-macros-0:1.24.0-1 100% | 381.7 KiB/s | 58.0 KiB | 00m00s [ 5/29] python3-typing-extensions-0:4.1 100% | 1.4 MiB/s | 114.1 KiB | 00m00s [ 6/29] python3-sortedcontainers-0:2.4. 100% | 2.0 MiB/s | 63.9 KiB | 00m00s [ 7/29] python3-iniconfig-0:2.1.0-3.fc4 100% | 846.9 KiB/s | 22.9 KiB | 00m00s [ 8/29] python3-packaging-0:25.0-7.fc43 100% | 2.9 MiB/s | 161.5 KiB | 00m00s [ 9/29] python3-devel-0:3.14.8-1.fc43.x 100% | 1.7 MiB/s | 439.4 KiB | 00m00s [10/29] python3-pluggy-0:1.6.0-4.fc43.n 100% | 2.1 MiB/s | 61.5 KiB | 00m00s [11/29] python-rpm-macros-0:3.14-5.fc43 100% | 640.6 KiB/s | 17.3 KiB | 00m00s [12/29] python3-rpm-macros-0:3.14-5.fc4 100% | 470.5 KiB/s | 11.8 KiB | 00m00s [13/29] python3-numpy-f2py-1:2.3.5-1.fc 100% | 7.2 MiB/s | 510.8 KiB | 00m00s [14/29] python3-numpy-1:2.3.5-1.fc43.x8 100% | 20.6 MiB/s | 8.2 MiB | 00m00s [15/29] mpdecimal-0:4.0.1-2.fc43.x86_64 100% | 2.9 MiB/s | 97.1 KiB | 00m00s [16/29] python3-0:3.14.8-1.fc43.x86_64 100% | 1.0 MiB/s | 29.4 KiB | 00m00s [17/29] expat-0:2.9.0-1.fc43.x86_64 100% | 4.1 MiB/s | 131.1 KiB | 00m00s [18/29] flexiblas-netlib-0:3.5.0-1.fc43 100% | 68.2 MiB/s | 4.5 MiB | 00m00s [19/29] flexiblas-0:3.5.0-1.fc43.x86_64 100% | 339.4 KiB/s | 23.1 KiB | 00m00s [20/29] flexiblas-openblas-openmp-0:3.5 100% | 507.6 KiB/s | 17.3 KiB | 00m00s [21/29] python3-libs-0:3.14.8-1.fc43.x8 100% | 37.8 MiB/s | 9.9 MiB | 00m00s [22/29] tzdata-0:2026e-1.fc43.noarch 100% | 23.2 MiB/s | 713.0 KiB | 00m00s [23/29] python-pip-wheel-0:25.1.1-19.fc 100% | 19.1 MiB/s | 1.2 MiB | 00m00s [24/29] libgfortran-0:15.3.1-1.fc43.x86 100% | 30.7 MiB/s | 973.6 KiB | 00m00s [25/29] libquadmath-0:15.3.1-1.fc43.x86 100% | 4.4 MiB/s | 213.7 KiB | 00m00s [26/29] openblas-openmp-0:0.3.29-2.fc43 100% | 36.7 MiB/s | 5.4 MiB | 00m00s [27/29] python3-attrs-0:25.4.0-1.fc43.n 100% | 2.8 MiB/s | 140.6 KiB | 00m00s [28/29] openblas-0:0.3.29-2.fc43.x86_64 100% | 1.6 MiB/s | 42.3 KiB | 00m00s [29/29] python3-rpm-generators-0:14-13. 100% | 1.1 MiB/s | 28.5 KiB | 00m00s -------------------------------------------------------------------------------- [29/29] Total 100% | 38.4 MiB/s | 36.6 MiB | 00m01s Running transaction [ 1/31] Verify package files 100% | 243.0 B/s | 29.0 B | 00m00s [ 2/31] Prepare transaction 100% | 763.0 B/s | 29.0 B | 00m00s [ 3/31] Installing python-rpm-macros-0: 100% | 0.0 B/s | 23.9 KiB | 00m00s [ 4/31] Installing python3-rpm-macros-0 100% | 0.0 B/s | 6.8 KiB | 00m00s [ 5/31] Installing libgfortran-0:15.3.1 100% | 477.7 MiB/s | 3.3 MiB | 00m00s [ 6/31] Installing pyproject-rpm-macros 100% | 167.4 MiB/s | 171.5 KiB | 00m00s [ 7/31] Installing openblas-0:0.3.29-2. 100% | 0.0 B/s | 113.5 KiB | 00m00s [ 8/31] Installing openblas-openmp-0:0. 100% | 661.9 MiB/s | 43.7 MiB | 00m00s [ 9/31] Installing libquadmath-0:15.3.1 100% | 311.7 MiB/s | 319.2 KiB | 00m00s [10/31] Installing flexiblas-netlib-0:3 100% | 358.9 MiB/s | 16.5 MiB | 00m00s [11/31] Installing flexiblas-0:3.5.0-1. 100% | 0.0 B/s | 39.2 KiB | 00m00s [12/31] Installing flexiblas-openblas-o 100% | 13.0 MiB/s | 40.1 KiB | 00m00s [13/31] Installing tzdata-0:2026e-1.fc4 100% | 76.5 MiB/s | 1.9 MiB | 00m00s [14/31] Installing python-pip-wheel-0:2 100% | 623.0 MiB/s | 1.2 MiB | 00m00s [15/31] Installing expat-0:2.9.0-1.fc43 100% | 27.6 MiB/s | 339.6 KiB | 00m00s [16/31] Installing mpdecimal-0:4.0.1-2. 100% | 42.7 MiB/s | 218.8 KiB | 00m00s [17/31] Installing python3-libs-0:3.14. 100% | 385.4 MiB/s | 43.9 MiB | 00m00s [18/31] Installing python3-0:3.14.8-1.f 100% | 2.0 MiB/s | 30.4 KiB | 00m00s [19/31] Installing python3-numpy-1:2.3. 100% | 385.9 MiB/s | 41.7 MiB | 00m00s [20/31] Installing python3-numpy-f2py-1 100% | 108.6 MiB/s | 2.2 MiB | 00m00s [21/31] Installing python3-packaging-0: 100% | 302.7 MiB/s | 619.9 KiB | 00m00s [22/31] Installing python3-rpm-generato 100% | 0.0 B/s | 82.9 KiB | 00m00s [23/31] Installing python3-sortedcontai 100% | 388.0 MiB/s | 397.3 KiB | 00m00s [24/31] Installing python3-iniconfig-0: 100% | 45.4 MiB/s | 46.5 KiB | 00m00s [25/31] Installing python3-pluggy-0:1.6 100% | 212.4 MiB/s | 217.5 KiB | 00m00s [26/31] Installing python3-attrs-0:25.4 100% | 146.8 MiB/s | 451.1 KiB | 00m00s [27/31] Installing python3-hypothesis-0 100% | 194.7 MiB/s | 5.1 MiB | 00m00s [28/31] Installing python3-pytest-0:8.3 100% | 411.2 MiB/s | 21.8 MiB | 00m00s [29/31] Installing python3-devel-0:3.14 100% | 105.6 MiB/s | 2.0 MiB | 00m00s [30/31] Installing python3-array-api-st 100% | 233.6 MiB/s | 717.6 KiB | 00m00s [31/31] Installing python3-typing-exten 100% | 13.5 MiB/s | 554.3 KiB | 00m00s Complete! Building target platforms: x86_64 Building for target x86_64 setting SOURCE_DATE_EPOCH=1793923200 Wrote: /builddir/build/SRPMS/python-array-api-extra-0.11.5-1.fc43.src.rpm Updating and loading repositories: Copr repository 100% | 5.8 KiB/s | 1.5 KiB | 00m00s fedora 100% | 46.6 KiB/s | 23.5 KiB | 00m01s updates 100% | 52.6 KiB/s | 18.6 KiB | 00m00s Repositories loaded. Package "pyproject-rpm-macros-1.24.0-1.fc43.noarch" is already installed. Package "python3-array-api-strict-2.0.1-7.fc43.noarch" is already installed. Package "python3-devel-3.14.8-1.fc43.x86_64" is already installed. Package "python3-hypothesis-6.123.0-7.fc43.noarch" is already installed. Package "python3-numpy-1:2.3.5-1.fc43.x86_64" is already installed. Package "python3-pytest-8.3.5-9.fc43.noarch" is already installed. Package "python3-typing-extensions-4.16.0-4.fc43.noarch" is already installed. Nothing to do. Finish: build setup for python-array-api-extra-0.11.5-1.fc43.src.rpm Start: rpmbuild python-array-api-extra-0.11.5-1.fc43.src.rpm Building target platforms: x86_64 Building for target x86_64 setting SOURCE_DATE_EPOCH=1793923200 Executing(%mkbuilddir): /bin/sh -e /var/tmp/rpm-tmp.UfkCv6 Executing(%prep): /bin/sh -e /var/tmp/rpm-tmp.FIEjdp + umask 022 + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + rm -rf array_api_extra-0.11.5 + /usr/lib/rpm/rpmuncompress -x /builddir/build/SOURCES/array_api_extra-0.11.5.tar.gz + STATUS=0 + '[' 0 -ne 0 ']' + cd array_api_extra-0.11.5 + /usr/bin/chmod -Rf a+rX,u+w,g-w,o-w . + RPM_EC=0 ++ jobs -p + exit 0 Executing(%generate_buildrequires): /bin/sh -e /var/tmp/rpm-tmp.dvYLyn + umask 022 + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + cd array_api_extra-0.11.5 + echo pyproject-rpm-macros + echo python3-devel + echo 'python3dist(packaging)' + echo 'python3dist(pip) >= 19' + '[' -f pyproject.toml ']' + echo '(python3dist(tomli) if python3-devel < 3.11)' + rm -rfv '*.dist-info/' + '[' -f /usr/bin/python3 ']' + mkdir -p /builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + echo -n + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + CXXFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + FFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + FCFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + VALAFLAGS=-g + RUSTFLAGS='-Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none -Cforce-frame-pointers=yes --cap-lints=warn' + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 ' + LT_SYS_LIBRARY_PATH=/usr/lib64: + CC=gcc + CXX=g++ + TMPDIR=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + RPM_TOXENV=py314 + FEDORA=43 + HOSTNAME=rpmbuild + /usr/bin/python3 -Bs /usr/lib/rpm/redhat/pyproject_buildrequires.py --generate-extras --python3_pkgversion 3 --wheeldir /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir --output /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-buildrequires --dep-overrides-file /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-dep-overrides -p Handling meson-python from build-system.requires Requirement not satisfied: meson-python Exiting dependency generation pass: build backend + cat /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-buildrequires + rm -rfv '*.dist-info/' + RPM_EC=0 ++ jobs -p + exit 0 Wrote: /builddir/build/SRPMS/python-array-api-extra-0.11.5-1.fc43.buildreqs.nosrc.rpm INFO: Going to install missing dynamic buildrequires Updating and loading repositories: Copr repository 100% | 7.1 KiB/s | 1.5 KiB | 00m00s fedora 100% | 55.5 KiB/s | 23.5 KiB | 00m00s updates 100% | 63.5 KiB/s | 18.6 KiB | 00m00s Repositories loaded. Package "pyproject-rpm-macros-1.24.0-1.fc43.noarch" is already installed. Package "python3-array-api-strict-2.0.1-7.fc43.noarch" is already installed. Package "python3-devel-3.14.8-1.fc43.x86_64" is already installed. Package "python3-hypothesis-6.123.0-7.fc43.noarch" is already installed. Package "python3-numpy-1:2.3.5-1.fc43.x86_64" is already installed. Package "python3-pytest-8.3.5-9.fc43.noarch" is already installed. Package "python3-typing-extensions-4.16.0-4.fc43.noarch" is already installed. Package "python3-packaging-25.0-7.fc43.noarch" is already installed. Package Arch Version Repository Size Installing: python3-meson-python noarch 0.18.0-6.fc43 fedora 326.5 KiB python3-pip noarch 25.1.1-19.fc43 updates 12.5 MiB Installing dependencies: meson noarch 1.8.5-3.fc43 updates 13.3 MiB ninja-build x86_64 1.13.1-4.fc43 fedora 480.7 KiB patchelf x86_64 0.18.0-9.fc43 fedora 287.6 KiB python3-pyproject-metadata noarch 0.9.1-5.fc43 fedora 193.6 KiB python3-setuptools noarch 78.1.1-15.fc43 fedora 9.0 MiB vim-filesystem noarch 2:9.2.1202-1.fc43 updates 40.0 B Transaction Summary: Installing: 8 packages Total size of inbound packages is 8 MiB. Need to download 8 MiB. After this operation, 36 MiB extra will be used (install 36 MiB, remove 0 B). [1/8] python3-meson-python-0:0.18.0-6.f 100% | 740.5 KiB/s | 93.3 KiB | 00m00s [2/8] patchelf-0:0.18.0-9.fc43.x86_64 100% | 912.1 KiB/s | 127.7 KiB | 00m00s [3/8] python3-pyproject-metadata-0:0.9. 100% | 1.9 MiB/s | 57.0 KiB | 00m00s [4/8] ninja-build-0:1.13.1-4.fc43.x86_6 100% | 3.3 MiB/s | 198.0 KiB | 00m00s [5/8] python3-setuptools-0:78.1.1-15.fc 100% | 21.0 MiB/s | 2.1 MiB | 00m00s [6/8] python3-pip-0:25.1.1-19.fc43.noar 100% | 6.6 MiB/s | 2.8 MiB | 00m00s [7/8] vim-filesystem-2:9.2.1202-1.fc43. 100% | 101.5 KiB/s | 15.5 KiB | 00m00s [8/8] meson-0:1.8.5-3.fc43.noarch 100% | 5.0 MiB/s | 2.4 MiB | 00m00s -------------------------------------------------------------------------------- [8/8] Total 100% | 12.4 MiB/s | 7.8 MiB | 00m01s Running transaction [ 1/10] Verify package files 100% | 320.0 B/s | 8.0 B | 00m00s [ 2/10] Prepare transaction 100% | 258.0 B/s | 8.0 B | 00m00s [ 3/10] Installing vim-filesystem-2:9.2 100% | 4.7 MiB/s | 4.8 KiB | 00m00s [ 4/10] Installing ninja-build-0:1.13.1 100% | 31.5 MiB/s | 483.8 KiB | 00m00s [ 5/10] Installing python3-setuptools-0 100% | 263.4 MiB/s | 9.2 MiB | 00m00s [ 6/10] Installing meson-0:1.8.5-3.fc43 100% | 297.8 MiB/s | 13.4 MiB | 00m00s [ 7/10] Installing python3-pyproject-me 100% | 194.4 MiB/s | 199.0 KiB | 00m00s [ 8/10] Installing patchelf-0:0.18.0-9. 100% | 23.6 MiB/s | 289.4 KiB | 00m00s [ 9/10] Installing python3-meson-python 100% | 81.2 MiB/s | 332.5 KiB | 00m00s [10/10] Installing python3-pip-0:25.1.1 100% | 185.3 MiB/s | 12.8 MiB | 00m00s Complete! Building target platforms: x86_64 Building for target x86_64 setting SOURCE_DATE_EPOCH=1793923200 Executing(%generate_buildrequires): /bin/sh -e /var/tmp/rpm-tmp.kvN800 + umask 022 + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + cd array_api_extra-0.11.5 + echo pyproject-rpm-macros + echo python3-devel + echo 'python3dist(packaging)' + echo 'python3dist(pip) >= 19' + '[' -f pyproject.toml ']' + echo '(python3dist(tomli) if python3-devel < 3.11)' + rm -rfv '*.dist-info/' + '[' -f /usr/bin/python3 ']' + mkdir -p /builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + echo -n + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + CXXFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + FFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + FCFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + VALAFLAGS=-g + RUSTFLAGS='-Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none -Cforce-frame-pointers=yes --cap-lints=warn' + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 ' + LT_SYS_LIBRARY_PATH=/usr/lib64: + CC=gcc + CXX=g++ + TMPDIR=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + RPM_TOXENV=py314 + FEDORA=43 + HOSTNAME=rpmbuild + /usr/bin/python3 -Bs /usr/lib/rpm/redhat/pyproject_buildrequires.py --generate-extras --python3_pkgversion 3 --wheeldir /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir --output /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-buildrequires --dep-overrides-file /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-dep-overrides -p Handling meson-python from build-system.requires Requirement satisfied: meson-python (installed: meson-python 0.18.0) Handling array-api-compat>=1.15.0,<2 from pyproject.toml generated metadata: [dependencies] (array-api-extra) Requirement not satisfied: array-api-compat<2,>=1.15.0 + cat /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-buildrequires + rm -rfv '*.dist-info/' + RPM_EC=0 ++ jobs -p + exit 0 Wrote: /builddir/build/SRPMS/python-array-api-extra-0.11.5-1.fc43.buildreqs.nosrc.rpm INFO: Going to install missing dynamic buildrequires Updating and loading repositories: Copr repository 100% | 10.1 KiB/s | 1.5 KiB | 00m00s fedora 100% | 68.6 KiB/s | 23.5 KiB | 00m00s updates 100% | 68.7 KiB/s | 18.6 KiB | 00m00s Repositories loaded. Package "pyproject-rpm-macros-1.24.0-1.fc43.noarch" is already installed. Package "python3-array-api-strict-2.0.1-7.fc43.noarch" is already installed. Package "python3-devel-3.14.8-1.fc43.x86_64" is already installed. Package "python3-hypothesis-6.123.0-7.fc43.noarch" is already installed. Package "python3-numpy-1:2.3.5-1.fc43.x86_64" is already installed. Package "python3-pytest-8.3.5-9.fc43.noarch" is already installed. Package "python3-typing-extensions-4.16.0-4.fc43.noarch" is already installed. Package "python3-meson-python-0.18.0-6.fc43.noarch" is already installed. Package "python3-packaging-25.0-7.fc43.noarch" is already installed. Package "python3-pip-25.1.1-19.fc43.noarch" is already installed. Package Arch Version Repository Size Installing: python3-array-api-compat noarch 1.15.0-1.fc43 copr_base 506.6 KiB Transaction Summary: Installing: 1 package Total size of inbound packages is 123 KiB. Need to download 123 KiB. After this operation, 507 KiB extra will be used (install 507 KiB, remove 0 B). [1/1] python3-array-api-compat-0:1.15.0 100% | 433.7 KiB/s | 123.2 KiB | 00m00s -------------------------------------------------------------------------------- [1/1] Total 100% | 432.2 KiB/s | 123.2 KiB | 00m00s Running transaction [1/3] Verify package files 100% | 0.0 B/s | 1.0 B | 00m00s [2/3] Prepare transaction 100% | 166.0 B/s | 1.0 B | 00m00s [3/3] Installing python3-array-api-comp 100% | 47.0 MiB/s | 529.3 KiB | 00m00s Warning: skipped OpenPGP checks for 1 package from repository: copr_base Complete! Building target platforms: x86_64 Building for target x86_64 setting SOURCE_DATE_EPOCH=1793923200 Executing(%generate_buildrequires): /bin/sh -e /var/tmp/rpm-tmp.6h4QXc + umask 022 + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + cd array_api_extra-0.11.5 + echo pyproject-rpm-macros + echo python3-devel + echo 'python3dist(packaging)' + echo 'python3dist(pip) >= 19' + '[' -f pyproject.toml ']' + echo '(python3dist(tomli) if python3-devel < 3.11)' + rm -rfv '*.dist-info/' + '[' -f /usr/bin/python3 ']' + mkdir -p /builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + echo -n + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + CXXFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + FFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + FCFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + VALAFLAGS=-g + RUSTFLAGS='-Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none -Cforce-frame-pointers=yes --cap-lints=warn' + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 ' + LT_SYS_LIBRARY_PATH=/usr/lib64: + CC=gcc + CXX=g++ + TMPDIR=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + RPM_TOXENV=py314 + FEDORA=43 + HOSTNAME=rpmbuild + /usr/bin/python3 -Bs /usr/lib/rpm/redhat/pyproject_buildrequires.py --generate-extras --python3_pkgversion 3 --wheeldir /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir --output /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-buildrequires --dep-overrides-file /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-dep-overrides -p Handling meson-python from build-system.requires Requirement satisfied: meson-python (installed: meson-python 0.18.0) Handling array-api-compat>=1.15.0,<2 from pyproject.toml generated metadata: [dependencies] (array-api-extra) Requirement satisfied: array-api-compat<2,>=1.15.0 (installed: array-api-compat 1.15.0) + cat /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-buildrequires + rm -rfv '*.dist-info/' + RPM_EC=0 ++ jobs -p + exit 0 Wrote: /builddir/build/SRPMS/python-array-api-extra-0.11.5-1.fc43.buildreqs.nosrc.rpm INFO: Going to install missing dynamic buildrequires Updating and loading repositories: Copr repository 100% | 13.3 KiB/s | 1.5 KiB | 00m00s fedora 100% | 95.3 KiB/s | 23.5 KiB | 00m00s updates 100% | 106.3 KiB/s | 18.6 KiB | 00m00s Repositories loaded. Package "pyproject-rpm-macros-1.24.0-1.fc43.noarch" is already installed. Package "python3-array-api-strict-2.0.1-7.fc43.noarch" is already installed. Package "python3-devel-3.14.8-1.fc43.x86_64" is already installed. Package "python3-hypothesis-6.123.0-7.fc43.noarch" is already installed. Package "python3-numpy-1:2.3.5-1.fc43.x86_64" is already installed. Package "python3-pytest-8.3.5-9.fc43.noarch" is already installed. Package "python3-typing-extensions-4.16.0-4.fc43.noarch" is already installed. Package "python3-meson-python-0.18.0-6.fc43.noarch" is already installed. Package "python3-packaging-25.0-7.fc43.noarch" is already installed. Package "python3-pip-25.1.1-19.fc43.noarch" is already installed. Nothing to do. Building target platforms: x86_64 Building for target x86_64 setting SOURCE_DATE_EPOCH=1793923200 Executing(%generate_buildrequires): /bin/sh -e /var/tmp/rpm-tmp.QSeOCD + umask 022 + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + cd array_api_extra-0.11.5 + echo pyproject-rpm-macros + echo python3-devel + echo 'python3dist(packaging)' + echo 'python3dist(pip) >= 19' + '[' -f pyproject.toml ']' + echo '(python3dist(tomli) if python3-devel < 3.11)' + rm -rfv '*.dist-info/' + '[' -f /usr/bin/python3 ']' + mkdir -p /builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + echo -n + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + CXXFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + FFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + FCFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + VALAFLAGS=-g + RUSTFLAGS='-Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none -Cforce-frame-pointers=yes --cap-lints=warn' + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 ' + LT_SYS_LIBRARY_PATH=/usr/lib64: + CC=gcc + CXX=g++ + TMPDIR=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + RPM_TOXENV=py314 + FEDORA=43 + HOSTNAME=rpmbuild + /usr/bin/python3 -Bs /usr/lib/rpm/redhat/pyproject_buildrequires.py --generate-extras --python3_pkgversion 3 --wheeldir /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir --output /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-buildrequires --dep-overrides-file /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-dep-overrides -p Handling meson-python from build-system.requires Requirement satisfied: meson-python (installed: meson-python 0.18.0) Handling array-api-compat>=1.15.0,<2 from pyproject.toml generated metadata: [dependencies] (array-api-extra) Requirement satisfied: array-api-compat<2,>=1.15.0 (installed: array-api-compat 1.15.0) + cat /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-buildrequires + rm -rfv '*.dist-info/' + RPM_EC=0 ++ jobs -p + exit 0 Executing(%build): /bin/sh -e /var/tmp/rpm-tmp.ZjT7mh + umask 022 + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + export CFLAGS + CXXFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + export CXXFLAGS + FFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + export FFLAGS + FCFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + export FCFLAGS + VALAFLAGS=-g + export VALAFLAGS + RUSTFLAGS='-Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none -Cforce-frame-pointers=yes -Clink-arg=-specs=/usr/lib/rpm/redhat/redhat-package-notes --cap-lints=warn' + export RUSTFLAGS + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 -specs=/usr/lib/rpm/redhat/redhat-package-notes ' + export LDFLAGS + LT_SYS_LIBRARY_PATH=/usr/lib64: + export LT_SYS_LIBRARY_PATH + CC=gcc + export CC + CXX=g++ + export CXX + cd array_api_extra-0.11.5 + mkdir -p /builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + CXXFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + FFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + FCFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + VALAFLAGS=-g + RUSTFLAGS='-Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none -Cforce-frame-pointers=yes -Clink-arg=-specs=/usr/lib/rpm/redhat/redhat-package-notes --cap-lints=warn' + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 -specs=/usr/lib/rpm/redhat/redhat-package-notes ' + LT_SYS_LIBRARY_PATH=/usr/lib64: + CC=gcc + CXX=g++ + TMPDIR=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + /usr/bin/python3 -Bs /usr/lib/rpm/redhat/pyproject_wheel.py /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir Processing /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5 Preparing metadata (pyproject.toml): started Running command Preparing metadata (pyproject.toml) + meson setup /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5 /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/.mesonpy-wnze733h -Dbuildtype=release -Db_ndebug=if-release -Db_vscrt=md --native-file=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/.mesonpy-wnze733h/meson-python-native-file.ini The Meson build system Version: 1.8.5 Source dir: /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5 Build dir: /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/.mesonpy-wnze733h Build type: native build Project name: array-api-extra Project version: 0.11.5 Host machine cpu family: x86_64 Host machine cpu: x86_64 Program python found: YES (/usr/bin/python3) Build targets in project: 0 array-api-extra 0.11.5 User defined options Native files: /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/.mesonpy-wnze733h/meson-python-native-file.ini b_ndebug : if-release b_vscrt : md buildtype : release Found ninja-1.13.1 at /usr/bin/ninja + /usr/bin/ninja ninja: no work to do. [1/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/__init__.py [2/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_at.py [3/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_creation.py [4/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_elementwise.py [5/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_indexing.py [6/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lazy.py [7/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_linalg.py [8/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_manipulation.py [9/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_searching.py [10/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_set.py [11/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_sorting.py [12/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_statistical.py [13/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/py.typed [14/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/__init__.py [15/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_creation.py [16/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_elementwise.py [17/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_indexing.py [18/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_inspection.py [19/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_linalg.py [20/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_manipulation.py [21/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_searching.py [22/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_set.py [23/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_sorting.py [24/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_agnostic/_statistical.py [25/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/testing/__init__.py [26/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/testing/_testing.py [27/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lib/__init__.py [28/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lib/_backends.py [29/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lib/_compat.py [30/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lib/_compat.pyi [31/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lib/_helpers.py [32/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lib/_testing.py [33/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lib/_typing.py [34/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/src/array_api_extra/_lib/_typing.pyi [35/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/__init__.py [36/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/__init__.py [37/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/conftest.py [38/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_at.py [39/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_creation.py [40/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_deprecation.py [41/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_elementwise.py [42/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_helpers.py [43/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_indexing.py [44/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_inspection.py [45/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_lazy.py [46/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_linalg.py [47/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_manipulation.py [48/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_public_api.py [49/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_searching.py [50/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_set.py [51/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_sorting.py [52/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_statistical.py [53/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_testing.py [54/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/main/test_version.py [55/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/run_deps/__init__.py [56/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/run_deps/test_run_deps.py [57/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/vendoring/__init__.py [58/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/vendoring/_array_api_compat_vendor.py [59/59] /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5/tests/vendoring/test_vendor.py Preparing metadata (pyproject.toml): finished with status 'done' Building wheels for collected packages: array-api-extra Building wheel for array-api-extra (pyproject.toml): started Running command Building wheel for array-api-extra (pyproject.toml) Building wheel for array-api-extra (pyproject.toml): finished with status 'done' Created wheel for array-api-extra: filename=array_api_extra-0.11.5-py3-none-any.whl size=119896 sha256=183f380a868929654a9ebda5a580569e379f320920e0f6084923e6ca0d1e8a49 Stored in directory: /builddir/.cache/pip/wheels/cc/dc/96/234ab57fef0e5fb0da4623334b61afaafe685d977f56dabfef Successfully built array-api-extra + RPM_EC=0 ++ jobs -p + exit 0 Executing(%install): /bin/sh -e /var/tmp/rpm-tmp.hQ6ag6 + umask 022 + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + '[' /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT '!=' / ']' + rm -rf /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT ++ dirname /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT + mkdir -p /builddir/build/BUILD/python-array-api-extra-0.11.5-build + mkdir /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + export CFLAGS + CXXFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + export CXXFLAGS + FFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + export FFLAGS + FCFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + export FCFLAGS + VALAFLAGS=-g + export VALAFLAGS + RUSTFLAGS='-Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none -Cforce-frame-pointers=yes -Clink-arg=-specs=/usr/lib/rpm/redhat/redhat-package-notes --cap-lints=warn' + export RUSTFLAGS + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 -specs=/usr/lib/rpm/redhat/redhat-package-notes ' + export LDFLAGS + LT_SYS_LIBRARY_PATH=/usr/lib64: + export LT_SYS_LIBRARY_PATH + CC=gcc + export CC + CXX=g++ + export CXX + cd array_api_extra-0.11.5 ++ ls /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir/array_api_extra-0.11.5-py3-none-any.whl ++ xargs basename --multiple ++ sed -E 's/([^-]+)-([^-]+)-.+\.whl/\1==\2/' + specifier=array_api_extra==0.11.5 + '[' -z array_api_extra==0.11.5 ']' + PIP_CONFIG_FILE=/dev/null + TMPDIR=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir + /usr/bin/python3 -m pip install --root /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT --prefix /usr --no-deps --disable-pip-version-check --progress-bar off --verbose --ignore-installed --no-warn-script-location --no-index --no-cache-dir --find-links /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir array_api_extra==0.11.5 Using pip 25.1.1 from /usr/lib/python3.14/site-packages/pip (python 3.14) Looking in links: /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir Processing /builddir/build/BUILD/python-array-api-extra-0.11.5-build/pyproject-wheeldir/array_api_extra-0.11.5-py3-none-any.whl Installing collected packages: array_api_extra Successfully installed array_api_extra-0.11.5 + '[' -d /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/bin ']' + rm -f /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-ghost-distinfo + site_dirs=() + '[' -d /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages ']' + site_dirs+=("/usr/lib/python3.14/site-packages") + '[' /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib64/python3.14/site-packages '!=' /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages ']' + '[' -d /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib64/python3.14/site-packages ']' + for site_dir in ${site_dirs[@]} + for distinfo in /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT$site_dir/*.dist-info + echo '%ghost %dir /usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info' +++ basename /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info ++ PYTHONPATH=/usr/lib/rpm/redhat ++ /usr/bin/python3 -sBc 'import sys; from pyproject_save_files import canonical_name_from_distinfo; print(canonical_name_from_distinfo(sys.argv[1]))' array_api_extra-0.11.5.dist-info + _normalized=array-api-extra + echo '%ghost %dir /usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info' + sed -i s/pip/rpm/ /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info/INSTALLER + PYTHONPATH=/usr/lib/rpm/redhat + /usr/bin/python3 -B /usr/lib/rpm/redhat/pyproject_preprocess_record.py --buildroot /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT --record /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info/RECORD --output /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-record + rm -fv /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info/RECORD removed '/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info/RECORD' + rm -fv /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info/REQUESTED removed '/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra-0.11.5.dist-info/REQUESTED' + '[' -f /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-dep-overrides ']' ++ wc -l /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-ghost-distinfo ++ cut -f1 '-d ' + lines=1 + '[' 1 -ne 1 ']' + RPM_FILES_ESCAPE=4.19 + /usr/bin/python3 /usr/lib/rpm/redhat/pyproject_save_files.py --output-files /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-files --output-modules /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-modules --buildroot /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT --sitelib /usr/lib/python3.14/site-packages --sitearch /usr/lib64/python3.14/site-packages --python-version 3.14 --pyproject-record /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-record --prefix /usr array_api_extra + /usr/lib/rpm/check-buildroot + /usr/lib/rpm/redhat/brp-ldconfig + /usr/lib/rpm/brp-compress + /usr/lib/rpm/brp-strip /usr/bin/strip + /usr/lib/rpm/brp-strip-comment-note /usr/bin/strip /usr/bin/objdump + /usr/lib/rpm/redhat/brp-strip-lto /usr/bin/strip + /usr/lib/rpm/check-rpaths + /usr/lib/rpm/redhat/brp-mangle-shebangs + /usr/lib/rpm/brp-remove-la-files + /usr/lib/rpm/redhat/brp-python-rpm-in-distinfo + env /usr/lib/rpm/redhat/brp-python-bytecompile '' 1 0 -j4 Bytecompiling .py files below /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14 using python3.14 + /usr/lib/rpm/redhat/brp-python-hardlink + /usr/bin/add-determinism --brp -j4 /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_sorting.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_set.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_statistical.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_searching.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_indexing.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_manipulation.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_elementwise.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_linalg.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_creation.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_lazy.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/__init__.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/__pycache__/__init__.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/vendoring/__pycache__/test_vendor.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_lazy.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/vendoring/__pycache__/_array_api_compat_vendor.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/vendoring/__pycache__/test_vendor.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/vendoring/__pycache__/__init__.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/run_deps/__pycache__/__init__.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/run_deps/__pycache__/test_run_deps.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_inspection.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_at.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_inspection.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_version.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/__pycache__/_at.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_version.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_indexing.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_indexing.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_testing.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_public_api.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_statistical.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_public_api.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_helpers.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_statistical.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_manipulation.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_sorting.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_sorting.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_helpers.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_testing.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_linalg.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_deprecation.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_set.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_linalg.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_set.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_manipulation.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_creation.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_searching.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_creation.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_lazy.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_at.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/__init__.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_elementwise.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_at.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_lazy.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_lib/__pycache__/_typing.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_lib/__pycache__/_testing.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/conftest.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_lib/__pycache__/_compat.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/conftest.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_lib/__pycache__/_backends.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_lib/__pycache__/__init__.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_lib/__pycache__/_helpers.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/testing/__pycache__/_testing.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/testing/__pycache__/__init__.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_lib/__pycache__/_helpers.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_statistical.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_manipulation.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/tests/main/__pycache__/test_elementwise.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/testing/__pycache__/_testing.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_sorting.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_linalg.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_inspection.cpython-314.opt-1.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_inspection.cpython-314.pyc: replacing with normalized version /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_set.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_searching.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_indexing.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_elementwise.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/__init__.cpython-314.opt-1.pyc: rewriting with normalized contents /builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages/array_api_extra/_agnostic/__pycache__/_creation.cpython-314.opt-1.pyc: rewriting with normalized contents Scanned 23 directories and 175 files, processed 78 inodes, 78 modified (44 replaced + 34 rewritten), 0 unsupported format, 0 errors Executing(%check): /bin/sh -e /var/tmp/rpm-tmp.xQAiat + umask 022 + cd /builddir/build/BUILD/python-array-api-extra-0.11.5-build + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + export CFLAGS + CXXFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + export CXXFLAGS + FFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + export FFLAGS + FCFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -I/usr/lib64/gfortran/modules ' + export FCFLAGS + VALAFLAGS=-g + export VALAFLAGS + RUSTFLAGS='-Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none -Cforce-frame-pointers=yes -Clink-arg=-specs=/usr/lib/rpm/redhat/redhat-package-notes --cap-lints=warn' + export RUSTFLAGS + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 -specs=/usr/lib/rpm/redhat/redhat-package-notes ' + export LDFLAGS + LT_SYS_LIBRARY_PATH=/usr/lib64: + export LT_SYS_LIBRARY_PATH + CC=gcc + export CC + CXX=g++ + export CXX + cd array_api_extra-0.11.5 + '[' '!' -f /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-modules ']' + PATH=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/bin:/usr/bin:/bin:/usr/sbin:/sbin + PYTHONPATH=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib64/python3.14/site-packages:/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages + _PYTHONSITE=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib64/python3.14/site-packages:/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages + PYTHONDONTWRITEBYTECODE=1 + /usr/bin/python3 -sP /usr/lib/rpm/redhat/import_all_modules.py -f /builddir/build/BUILD/python-array-api-extra-0.11.5-build/python-array-api-extra-0.11.5-1.fc43.x86_64-pyproject-modules -e 'array_api_extra.tests*' Check import: array_api_extra Check import: array_api_extra.testing + CFLAGS='-O2 -flto=auto -ffat-lto-objects -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer ' + LDFLAGS='-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 -specs=/usr/lib/rpm/redhat/redhat-package-notes ' + PATH=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/bin:/usr/bin:/bin:/usr/sbin:/sbin + PYTHONPATH=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib64/python3.14/site-packages:/builddir/build/BUILD/python-array-api-extra-0.11.5-build/BUILDROOT/usr/lib/python3.14/site-packages + PYTHONDONTWRITEBYTECODE=1 + PYTEST_ADDOPTS=' --ignore=/builddir/build/BUILD/python-array-api-extra-0.11.5-build/.pyproject-builddir' + PYTEST_XDIST_AUTO_NUM_WORKERS=4 + /usr/bin/pytest -p no:cacheprovider tests --ignore=tests/vendoring -o addopts=-ra -W ignore::pytest.PytestUnknownMarkWarning ============================= test session starts ============================== platform linux -- Python 3.14.8, pytest-8.3.5, pluggy-1.6.0 rootdir: /builddir/build/BUILD/python-array-api-extra-0.11.5-build/array_api_extra-0.11.5 configfile: pyproject.toml plugins: hypothesis-6.123.0 collected 15283 items tests/main/test_at.py EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 0%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 0%] EEEEEEEEEEEEEEEEEEEEEEssssssssssssssssssssssssssssssssssssssssssssssssss [ 1%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 1%] ssssssssssssssssssssss.................................................. [ 2%] ........................................................................ [ 2%] ........................................................................ [ 3%] ........................................................................ [ 3%] ......................ssssssssssssssssssssssssssssssssssssssssssssssssss [ 4%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 4%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 5%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 5%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 5%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 6%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 6%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 7%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 7%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 8%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 8%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 9%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 9%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 10%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 10%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 11%] ssssssssssssssssssssssEEEEEEEEssssssss................ssssssssssssssssss [ 11%] ssssssssssssssssssssssssssssssssssssssssssssss...EEEEEEEEEEEEEEEEEEEEEEE [ 12%] EEEEEEEEEssssssssssssssssssssssssssssssss............................... [ 12%] .................................sssssssssssssssssssssssssssssssssssssss [ 13%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 13%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 13%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 14%] sEs..ssssssssEEss....ssssssssssssssssEEEEssss.s.s.s.ssssssssssssssssssss [ 14%] sssssssssssssEs..ssssssss [ 15%] tests/main/test_creation.py EEEEEEEEEEEEEEEEEEEEEEEE.................... [ 15%] ....ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 15%] ssssssssssssssssssssssssssssEE..ssssssssEE..ssssssssEs..ssssssssEE..ssss [ 16%] ssssEEEEEEEEEEEEEEEE................ssssssssssssssssssssssssssssssssssss [ 16%] ssssssssssssssssssssssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssss [ 17%] ssssEE..ssssssssEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 17%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 18%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE...................................... [ 18%] ........................................................................ [ 19%] ....................................................ssssssssssssssssssss [ 19%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 20%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 20%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 21%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 21%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 21%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 22%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 22%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 23%] ssssssssssssssssssssssssssssssssssssssssssssssssssssEE..ssssssssEEEEEEEE [ 23%] EEEEEEEEEEEE....................ssssssssssssssssssssssssssssssssssssssss [ 24%] ssssssssssssssssssssssssssssssssssssssssEE..ssssssssEE..ssssssss [ 24%] tests/main/test_deprecation.py EE..ssssssssEE..ssssssss [ 24%] tests/main/test_elementwise.py EE..ssssssssEE..ssssssssEE..ssssssssEE..s [ 25%] sssssssEE..ssssssssEEEEEEEE........ssssssssssssssssssssssssssssssssEE..s [ 25%] sssssssEs..ssssssssEE..ssssssssEE.sssssssssEE..ssssssssEE..ssssssssEEEEE [ 26%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 26%] EEEEEEEEEEEEEEE......................................................... [ 27%] ...................................sssssssssssssssssssssssssssssssssssss [ 27%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 28%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 28%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 28%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 29%] sssssssssssssssssssssssssssssssssssssssssssEEEE....ssssssssssssssssEE..s [ 29%] sssssssEE..ssssssssEEEEEE......ssssssssssssssssssssssssEEEE....sssssssss [ 30%] sssssssEE..ssssssssEE..ssssssssEs..ssssssssEs..ssssssssEE.sssssssss.EE.. [ 30%] ssssssssEEEE....ssssssssssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE.. [ 31%] ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEEEEEEEEEE...... [ 31%] ....ssssssssssssssssssssssssssssssssssssssssEE..ssssssssEE..ssssssssEEEE [ 32%] ....ssssssssssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE.. [ 32%] ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE.. [ 33%] ssssssssEEEE....ssssssssssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE.. [ 33%] ssssssssEEEE....ssssssssssssssssEE..ssssssssEE..ssssssssEE..ssssssss [ 34%] tests/main/test_helpers.py EEss....ssssssssssssssssEs..ssssssssEs.ssssss [ 34%] sssEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE................................sssss [ 34%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 35%] sssssssssssssssssssssssssssssssssssssssssssssssssssEE..ssssssssEEEEEEEEE [ 35%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 36%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 36%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 37%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 37%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE............................... [ 38%] ........................................................................ [ 38%] ........................................................................ [ 39%] ........................................................................ [ 39%] ........................................................................ [ 40%] ...................sssssssssssssssssssssssssssssssssssssssssssssssssssss [ 40%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 40%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 41%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 41%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 42%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 42%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 43%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 43%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 44%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 44%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 45%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 45%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 46%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 46%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 47%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 47%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 48%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 48%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 49%] sss..........Es..ssssssssEE.ssssssssssEE..ssssssssEE..ssssssssEE..ssssss [ 49%] ss.......sssssssss [ 49%] tests/main/test_indexing.py EE..ssssssssEEEEEEEEEEEEEEEEEEEEEEEE........ [ 49%] ................ssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 50%] ssssssssssssssssssssssssssssssssssssssssEE..ssssssssEE..ssssssssEE..ssss [ 50%] ssssEEEE....ssssssssssssssssEEEEEEEEEEEEEEEEEEEE....................ssss [ 51%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 51%] ssssEEEE....ssssssssssssssssEEEE....ssssssssssssssssEEEE....ssssssssssss [ 52%] ssssEEEE....ssssssssssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssss [ 52%] ssssEE.sssssssss [ 52%] tests/main/test_inspection.py EE..ssssssssEE..ssssssssEE..ssssssssss [ 53%] tests/main/test_lazy.py EEEEEEEEEEEEEEEE................ssssssssssssssss [ 53%] ssssssssssssssssssssssssssssssssssssssssssssssssEEEE....ssssssssssssssss [ 53%] EEEE....ssssssssssssssssEEEE....ssssssssssssssssssssEE..ssssssssEs..ssss [ 54%] ssssEsEE....ssssssssssssssssEE..ssssssssEE..ssssssssEEEE....ssssssssssss [ 54%] ssssEE..ssssssss. [ 54%] tests/main/test_linalg.py EE..ssssssssEE..ssssssssEEEEEEEEEEEE.......... [ 55%] ..ssssssssssssssssssssssssssssssssssssssssssssssssEE..ssssssss.EE..sssss [ 55%] sssEE..ssssssss [ 55%] tests/main/test_manipulation.py EE..ssssssssEEEEEEEEEE..........ssssssss [ 56%] ssssssssssssssssssssssssssssssssEE..ssssssssEEEEEEEEEE..........ssssssss [ 56%] ssssssssssssssssssssssssssssssssEE..ssssssssEEEEEEEEEE..........ssssssss [ 56%] ssssssssssssssssssssssssssssssssEE..ssssssssEEEEEEEEEE..........ssssssss [ 57%] ssssssssssssssssssssssssssssssssEE..ssssssss..EE.sssssssss.............. [ 57%] .........EE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE. [ 58%] .ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE. [ 58%] .ssssssssEE..ssssssssEE..ssssssssEE..ssssssss [ 59%] tests/main/test_public_api.py . [ 59%] tests/main/test_searching.py EE..ssssssssEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 59%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 59%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 60%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 60%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 61%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 61%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 62%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 62%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 63%] EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE [ 63%] EEEEEEEEEEEEEEEEEEEEEEEEE............................................... [ 64%] ........................................................................ [ 64%] ........................................................................ [ 65%] .............ssssssssssssssss............................sssssssssssssss [ 65%] s............................ssssssssssssssss........................... [ 66%] .ssssssssssssssss....................................................... [ 66%] ........................................................................ [ 66%] ........................................................................ [ 67%] .....ssssssssssssssss............................ssssssssssssssss....... [ 67%] .....................ssssssssssssssss............................sssssss [ 68%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 68%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 69%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 69%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 70%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 70%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 71%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 71%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 72%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 72%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 73%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 73%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 74%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 74%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 74%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 75%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 75%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 76%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 76%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 77%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 77%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 78%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 78%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 79%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 79%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 80%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 80%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 81%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 81%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 82%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 82%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 82%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 83%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 83%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 84%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 84%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 85%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 85%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 86%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 86%] sssssssssssssssss [ 86%] tests/main/test_set.py ss..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssE [ 87%] E..ssssssssEE..ssssssssEEEEEE......ssssssssssssssssssssssssEE..ssssssssE [ 87%] s..ssssssssEs..ssssssssEEEEEEEEEEEEEEEEEEssssssssssssssssss............. [ 88%] .......................sssssssssssssssssssssssssssssssssssssssssssssssss [ 88%] ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss [ 89%] sssssssssssssssssssssssEEss..ssssssssssssssssss..EEss....sssssssssssssss [ 89%] sEs.sssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEs..sssssss [ 90%] sEs..ssssssssEs..ssssssssEs..ssssssss [ 90%] tests/main/test_sorting.py EE..ssssssssEEEEEE......sssssssssssssssssssss [ 90%] sssEE..ssssssssEE..ssssssssEEEEEE......ssssssssssssssssssssssssEE..sssss [ 91%] sss [ 91%] tests/main/test_statistical.py EE..ssssssssEE..ssssssssEE..ssssssssEE..s [ 91%] sssssssEE..ssssssssEE..ssssssssEE.sssssssssEE..ssssssssEEEEEEEE........s [ 91%] sssssssssssssssssssssssssssssssEEEEEEEE........sssssssssssssssssssssssss [ 92%] sssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..s [ 92%] sssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..sssssssss.ssE [ 93%] E..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssEEEEEE......s [ 93%] sssssssssssssssssssssssEEEE....ssssssssssssssssEE..ssssssssEE..ssssssssE [ 94%] E..ssssssssEE..ssssssssEE..ssssssssEEEEEE......ssssssssssssssssssssssssE [ 94%] EEE....ssssssssssssssssEE..ssssssssEE..ssssssssEE..ssssssssEE..ssssssssE [ 95%] EEEEE......ssssssssssssssssssssssssEEEE....ssssssssssssssssEE..ssssssssE [ 95%] E..ssssssssEE..ssssssssEE..ssssssssEEEEEE......ssssssssssssssssssssssssE [ 96%] EEE....ssssssssssssssss [ 96%] tests/main/test_testing.py EE..ssssssssEE..ssssssssEEEE....sssssssssssss [ 96%] sssEEEEEEEE........ssssssssssssssssssssssssssssssssEEEEEEEEsssssssssssss [ 96%] sssssssssssssssssssssssssssEEEEEEEE........sssssssssssssssssssssssssssss [ 97%] sssEEEEEEEE........ssssssssssssssssssssssssssssssssEEEEEEEE........sssss [ 97%] sssssssssssssssssssssssssssEEEE....ssssssssssssssssEE..ssssssssEEEEssss. [ 98%] .......ssssssssssssssssssssssssssssssssEEEEEEEE........sssssssssssssssss [ 98%] sssssssssssssssEE..ssssssssEE..ssssssssEE..ssssssss.EEssssssssss.EE..sss [ 99%] sssssEE..ssssssssEE..ssssssssEE..ssssssssEE..sssssssssEE..sssssssssEE..s [ 99%] sssssssEE..sssssssssEE..ssssssss. [ 99%] tests/main/test_version.py . [ 99%] tests/run_deps/test_run_deps.py s [100%] ==================================== ERRORS ==================================== _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-set-40.0-expect_list0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-set-40.0-expect_list0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-set-40.0-expect_list0-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-add-40.0-expect_list1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-add-40.0-expect_list1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-add-40.0-expect_list1-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-subtract-100.0-expect_list2-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-subtract-100.0-expect_list2-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-subtract-100.0-expect_list2-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-multiply-2.0-expect_list3-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-multiply-2.0-expect_list3-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-multiply-2.0-expect_list3-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-divide-2.0-expect_list4-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-divide-2.0-expect_list4-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-divide-2.0-expect_list4-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-power-2.0-expect_list5-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-power-2.0-expect_list5-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-power-2.0-expect_list5-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-min-25.0-expect_list6-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-min-25.0-expect_list6-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-min-25.0-expect_list6-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-max-25.0-expect_list7-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-max-25.0-expect_list7-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-0-max-25.0-expect_list7-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-set-40.0-expect_list0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-set-40.0-expect_list0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-set-40.0-expect_list0-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-add-40.0-expect_list1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-add-40.0-expect_list1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-add-40.0-expect_list1-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-subtract-100.0-expect_list2-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-subtract-100.0-expect_list2-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-subtract-100.0-expect_list2-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-multiply-2.0-expect_list3-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-multiply-2.0-expect_list3-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-multiply-2.0-expect_list3-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-divide-2.0-expect_list4-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-divide-2.0-expect_list4-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-divide-2.0-expect_list4-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-power-2.0-expect_list5-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-power-2.0-expect_list5-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-power-2.0-expect_list5-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-min-25.0-expect_list6-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-min-25.0-expect_list6-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-min-25.0-expect_list6-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-max-25.0-expect_list7-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-max-25.0-expect_list7-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-1-1-max-25.0-expect_list7-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-set-40.0-expect_list0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-set-40.0-expect_list0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-set-40.0-expect_list0-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-add-40.0-expect_list1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-add-40.0-expect_list1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-add-40.0-expect_list1-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-subtract-100.0-expect_list2-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-subtract-100.0-expect_list2-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-subtract-100.0-expect_list2-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-multiply-2.0-expect_list3-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-multiply-2.0-expect_list3-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-multiply-2.0-expect_list3-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-divide-2.0-expect_list4-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-divide-2.0-expect_list4-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-divide-2.0-expect_list4-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-power-2.0-expect_list5-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-power-2.0-expect_list5-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-power-2.0-expect_list5-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-min-25.0-expect_list6-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-min-25.0-expect_list6-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-min-25.0-expect_list6-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-max-25.0-expect_list7-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-max-25.0-expect_list7-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-0-max-25.0-expect_list7-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-set-40.0-expect_list0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-set-40.0-expect_list0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-set-40.0-expect_list0-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-add-40.0-expect_list1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-add-40.0-expect_list1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-add-40.0-expect_list1-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-subtract-100.0-expect_list2-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-subtract-100.0-expect_list2-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-subtract-100.0-expect_list2-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-multiply-2.0-expect_list3-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-multiply-2.0-expect_list3-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-multiply-2.0-expect_list3-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-divide-2.0-expect_list4-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-divide-2.0-expect_list4-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-divide-2.0-expect_list4-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-power-2.0-expect_list5-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-power-2.0-expect_list5-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-power-2.0-expect_list5-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-min-25.0-expect_list6-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-min-25.0-expect_list6-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-min-25.0-expect_list6-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-max-25.0-expect_list7-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-max-25.0-expect_list7-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-1-1-max-25.0-expect_list7-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-set-40.0-expect_list0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-set-40.0-expect_list0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-set-40.0-expect_list0-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-add-40.0-expect_list1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-add-40.0-expect_list1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-add-40.0-expect_list1-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-subtract-100.0-expect_list2-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-subtract-100.0-expect_list2-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-subtract-100.0-expect_list2-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-multiply-2.0-expect_list3-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-multiply-2.0-expect_list3-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-multiply-2.0-expect_list3-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-divide-2.0-expect_list4-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-divide-2.0-expect_list4-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-divide-2.0-expect_list4-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-power-2.0-expect_list5-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-power-2.0-expect_list5-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-power-2.0-expect_list5-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-min-25.0-expect_list6-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-min-25.0-expect_list6-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-min-25.0-expect_list6-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-max-25.0-expect_list7-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-max-25.0-expect_list7-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-False-0-0-max-25.0-expect_list7-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-set-40.0-expect_list0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-set-40.0-expect_list0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-set-40.0-expect_list0-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-add-40.0-expect_list1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-add-40.0-expect_list1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-add-40.0-expect_list1-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-subtract-100.0-expect_list2-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-subtract-100.0-expect_list2-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-subtract-100.0-expect_list2-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-multiply-2.0-expect_list3-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-multiply-2.0-expect_list3-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-multiply-2.0-expect_list3-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-divide-2.0-expect_list4-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-divide-2.0-expect_list4-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-divide-2.0-expect_list4-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-power-2.0-expect_list5-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-power-2.0-expect_list5-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-power-2.0-expect_list5-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-min-25.0-expect_list6-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-min-25.0-expect_list6-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-min-25.0-expect_list6-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-max-25.0-expect_list7-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-max-25.0-expect_list7-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_update_ops[array_api_strict-True-0-0-max-25.0-expect_list7-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_copy_default[array_api_strict-set] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_copy_default[array_api_strict-add] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of test_copy_default[array_api_strict-subtract] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of test_copy_default[array_api_strict-multiply] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of test_copy_default[array_api_strict-divide] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of test_copy_default[array_api_strict-power] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_copy_default[array_api_strict-min] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_copy_default[array_api_strict-max] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-set-False-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-set-False-None] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-set-True-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-set-True-None] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-add-False-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-add-False-None] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-add-True-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-add-True-None] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-subtract-False-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-subtract-False-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-subtract-True-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-subtract-True-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-multiply-False-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-multiply-False-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-multiply-True-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-multiply-True-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-divide-False-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-divide-False-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-divide-True-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-divide-True-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-power-False-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-power-False-None] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-power-True-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_incompatible_dtype[array_api_strict-power-True-None] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-min-False-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-min-False-None] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-min-True-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-min-True-None] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-max-False-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-max-False-None] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-max-True-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_incompatible_dtype[array_api_strict-max-True-None] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of test_bool_mask_nd[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of test_no_inf_warnings[array_api_strict-False] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of test_no_inf_warnings[array_api_strict-True] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_gh134[array_api_strict-False-None] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_gh134[array_api_strict-False-False] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of test_gh134[array_api_strict-True-None] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_gh134[array_api_strict-True-False] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______________ ERROR at setup of test_device[array_api_strict] ________________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-1-0] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-1-1] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-1-2] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-1-3] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-3-0] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-3-1] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-3-2] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-3-3] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-10-0] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-10-1] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-10-2] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strict-10-3] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-1-0] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-1-1] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-1-2] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-1-3] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-3-0] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-3-1] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-3-2] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-3-3] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-10-0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-10-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-10-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestOneHot.test_dims_and_classes[array_api_strictest-10-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestOneHot.test_basic[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestOneHot.test_basic[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestOneHot.test_2d[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestOneHot.test_2d[array_api_strictest] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestOneHot.test_abstract_size[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestOneHot.test_out_of_bound[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_out_of_bound[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestOneHot.test_int_types[array_api_strict-int8] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_int_types[array_api_strict-int16] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_int_types[array_api_strict-int32] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_int_types[array_api_strict-int64] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_int_types[array_api_strict-uint8] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_int_types[array_api_strict-uint16] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_int_types[array_api_strict-uint32] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_int_types[array_api_strict-uint64] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestOneHot.test_int_types[array_api_strictest-int8] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestOneHot.test_int_types[array_api_strictest-int16] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestOneHot.test_int_types[array_api_strictest-int32] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestOneHot.test_int_types[array_api_strictest-int64] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestOneHot.test_int_types[array_api_strictest-uint8] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_int_types[array_api_strictest-uint16] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_int_types[array_api_strictest-uint32] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestOneHot.test_int_types[array_api_strictest-uint64] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestOneHot.test_custom_dtype[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestOneHot.test_custom_dtype[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestOneHot.test_axis[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestOneHot.test_axis[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestOneHot.test_non_integer[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestOneHot.test_non_integer[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestOneHot.test_device[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestOneHot.test_device[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestCreateDiagonal.test_1d_from_numpy[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_numpy[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-1-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-2-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-3-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-4-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-5-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-6-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-7-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-8-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strict-9-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-1-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-2-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-3-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-4-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-5-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-6-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-7-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-8-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-6] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-7] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestCreateDiagonal.test_1d_from_scipy[array_api_strictest-9-9] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_0d_raises[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_0d_raises[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape0] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape1] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape2] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape3] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape4] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape5] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape6] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape7] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape8] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strict-shape9] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape0] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape1] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape2] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape3] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape4] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape5] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape6] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape7] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape8] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCreateDiagonal.test_nd[array_api_strictest-shape9] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestCreateDiagonal.test_device[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCreateDiagonal.test_device[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestCreateDiagonal.test_xp[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestCreateDiagonal.test_xp[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestDeprecatedFunctions.test_broadcast_shapes[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestDeprecatedFunctions.test_broadcast_shapes[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestDeprecatedFunctions.test_expand_dims[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestDeprecatedFunctions.test_expand_dims[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestApplyWhere.test_f1_f2[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestApplyWhere.test_f1_f2[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestApplyWhere.test_fill_value[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestApplyWhere.test_fill_value[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestApplyWhere.test_args_tuple[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestApplyWhere.test_args_tuple[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestApplyWhere.test_broadcast[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestApplyWhere.test_broadcast[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestApplyWhere.test_dtype_propagation[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation_fill_value[array_api_strict-int32-int32-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation_fill_value[array_api_strict-int32-int32-fill_value_raw1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation_fill_value[array_api_strict-int8-int16-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation_fill_value[array_api_strict-int8-int16-fill_value_raw1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation_fill_value[array_api_strictest-int32-int32-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation_fill_value[array_api_strictest-int32-int32-fill_value_raw1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation_fill_value[array_api_strictest-int8-int16-3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dtype_propagation_fill_value[array_api_strictest-int8-int16-fill_value_raw1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dont_overwrite_fill_value[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestApplyWhere.test_dont_overwrite_fill_value[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestApplyWhere.test_dont_run_on_false[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestApplyWhere.test_bad_args[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestApplyWhere.test_bad_args[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestApplyWhere.test_xp[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestApplyWhere.test_xp[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestApplyWhere.test_device[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestApplyWhere.test_device[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestApplyWhere.test_hypothesis[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestApplyWhere.test_hypothesis[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-0.0-0.0-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-0.0-0.0-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-1.0-1.0-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-1.0-1.0-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-1.0-2.0-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-1.0-2.0-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strict-1.0--1.0-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-1.0--1.0-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-100.0-101.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-100.0-101.0-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_basic[array_api_strict-0-0-False] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_basic[array_api_strict-0-0-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_basic[array_api_strict-1-1-False] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_basic[array_api_strict-1-1-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_basic[array_api_strict-1-2-False] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_basic[array_api_strict-1-2-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestIsClose.test_basic[array_api_strict-1--1-False] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_basic[array_api_strict-1--1-True] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-(1+1j)-(1+1j)-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-(1+1j)-(1+1j)-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-(1+1j)-(1-1j)-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-(1+1j)-(1-1j)-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-inf-inf-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-inf-inf-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strict-inf-100.0-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strict-inf-100.0-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strict-inf--inf-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-inf--inf-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strict--inf--inf-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strict--inf--inf-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-nan-nan-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strict-nan-nan-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strict-nan-100.0-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strict-nan-100.0-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-1000000.0-1000001.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-1000000.0-1000001.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-1000000.0-1000100.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-1000000.0-1000100.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-1e-06-1.1e-06-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-1e-06-1.1e-06-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-1e-07-1.1e-07-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-1e-07-1.1e-07-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-(1000000+0j)-(1000000+1j)-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-(1000000+0j)-(1000000+1j)-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-(1000000+0j)-(1000000+100j)-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strict-(1000000+0j)-(1000000+100j)-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-0.0-0.0-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strictest-0.0-0.0-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1.0-1.0-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1.0-1.0-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1.0-2.0-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1.0-2.0-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1.0--1.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1.0--1.0-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-100.0-101.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-100.0-101.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strictest-0-0-False] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestIsClose.test_basic[array_api_strictest-0-0-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1-1-False] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1-1-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1-2-False] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1-2-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1--1-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1--1-True] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-(1+1j)-(1+1j)-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-(1+1j)-(1+1j)-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-(1+1j)-(1-1j)-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-(1+1j)-(1-1j)-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-inf-inf-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strictest-inf-inf-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-inf-100.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-inf-100.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-inf--inf-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-inf--inf-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest--inf--inf-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest--inf--inf-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-nan-nan-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_basic[array_api_strictest-nan-nan-True] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-nan-100.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-nan-100.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1000000.0-1000001.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1000000.0-1000001.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1000000.0-1000100.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1000000.0-1000100.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1e-06-1.1e-06-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1e-06-1.1e-06-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1e-07-1.1e-07-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-1e-07-1.1e-07-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-(1000000+0j)-(1000000+1j)-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-(1000000+0j)-(1000000+1j)-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-(1000000+0j)-(1000000+100j)-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_basic[array_api_strictest-(1000000+0j)-(1000000+100j)-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestIsClose.test_broadcast[array_api_strict-float32] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_broadcast[array_api_strict-int32] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_broadcast[array_api_strictest-float32] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_broadcast[array_api_strictest-int32] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestIsClose.test_some_inf[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestIsClose.test_some_inf[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestIsClose.test_equal_nan[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestIsClose.test_equal_nan[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestIsClose.test_tolerance[array_api_strict-float32] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_tolerance[array_api_strict-complex64] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_tolerance[array_api_strict-int32] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_tolerance[array_api_strictest-float32] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_tolerance[array_api_strictest-complex64] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_tolerance[array_api_strictest-int32] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_tolerance_integer_overflow[array_api_strict-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_tolerance_integer_overflow[array_api_strict-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_tolerance_integer_overflow[array_api_strictest-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_tolerance_integer_overflow[array_api_strictest-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestIsClose.test_very_small_numbers[array_api_strict] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsClose.test_very_small_numbers[array_api_strictest] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestIsClose.test_bool_dtype[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestIsClose.test_bool_dtype[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestIsClose.test_none_shape[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_none_shape_bool[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestIsClose.test_python_scalar[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestIsClose.test_python_scalar[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestIsClose.test_xp[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestIsClose.test_xp[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestIsClose.test_device[array_api_strict-True] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestIsClose.test_device[array_api_strict-False] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_device[array_api_strictest-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestIsClose.test_device[array_api_strictest-False] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_array_on_device_with_scalar[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsClose.test_array_on_device_with_scalar[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanToNum.test_bool[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanToNum.test_bool[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanToNum.test_scalar_pos_inf[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanToNum.test_scalar_pos_inf[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanToNum.test_scalar_neg_inf[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanToNum.test_scalar_neg_inf[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanToNum.test_scalar_nan[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanToNum.test_scalar_nan[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanToNum.test_real[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanToNum.test_real[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanToNum.test_complex[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanToNum.test_complex[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNanToNum.test_empty_array[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanToNum.test_empty_array[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strict-in_vals0-3-out_vals0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strict-in_vals1-3.0-out_vals1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strict-in_vals2-3-out_vals2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strict-in_vals3-3.0-out_vals3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strict-in_vals4-3.0-out_vals4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strictest-in_vals0-3-out_vals0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strictest-in_vals1-3.0-out_vals1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strictest-in_vals2-3-out_vals2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strictest-in_vals3-3.0-out_vals3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_success[array_api_strictest-in_vals4-3.0-out_vals4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanToNum.test_fill_value_failure[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanToNum.test_fill_value_failure[array_api_strictest] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestSinc.test_simple[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestSinc.test_simple[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestSinc.test_dtype[array_api_strict-0] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestSinc.test_dtype[array_api_strict-(1+3j)] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestSinc.test_dtype[array_api_strictest-0] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestSinc.test_dtype[array_api_strictest-(1+3j)] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____________ ERROR at setup of TestSinc.test_3d[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestSinc.test_3d[array_api_strictest] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestSinc.test_device[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestSinc.test_device[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____________ ERROR at setup of TestSinc.test_xp[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestSinc.test_xp[array_api_strictest] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestAngle.test_simple[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestAngle.test_simple[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestAngle.test_basic[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestAngle.test_basic[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestAngle.test_real[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestAngle.test_real[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestAngle.test_complex[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestAngle.test_complex[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestAngle.test_integral[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestAngle.test_integral[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestAngle.test_2d[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestAngle.test_2d[array_api_strictest] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestAngle.test_device[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestAngle.test_device[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestDeg2Rad.test_basic[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestDeg2Rad.test_basic[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDeg2Rad.test_integral[array_api_strict-int32] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDeg2Rad.test_integral[array_api_strict-int64] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestDeg2Rad.test_integral[array_api_strictest-int32] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestDeg2Rad.test_integral[array_api_strictest-int64] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestDeg2Rad.test_complex[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestDeg2Rad.test_complex[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestDeg2Rad.test_bool[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestDeg2Rad.test_bool[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestDeg2Rad.test_device[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestDeg2Rad.test_device[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestRad2Deg.test_basic[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestRad2Deg.test_basic[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestRad2Deg.test_integral[array_api_strict-int32] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestRad2Deg.test_integral[array_api_strict-int64] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestRad2Deg.test_integral[array_api_strictest-int32] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestRad2Deg.test_integral[array_api_strictest-int64] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestRad2Deg.test_complex[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestRad2Deg.test_complex[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestRad2Deg.test_bool[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestRad2Deg.test_bool[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestRad2Deg.test_device[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestRad2Deg.test_device[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIn1D.test_no_invert_assume_unique[array_api_strict-fast path] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIn1D.test_no_invert_assume_unique[array_api_strict-slow path] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestIn1D.test_device[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____________ ERROR at setup of TestIn1D.test_xp[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-bool-True-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-int16-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-uint8-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-float32-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-float64-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-complex64-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-complex128-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-float32-1.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-float64-1.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-complex64-1.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-complex128-1.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-complex64-1j-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-complex128-1j-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-bool-1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-int64-1.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strict-float64-1j-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-bool-True-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-int16-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-uint8-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-float32-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-float64-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-complex64-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-complex128-1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-float32-1.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-float64-1.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-complex64-1.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-complex128-1.0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-complex64-1j-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-complex128-1j-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-bool-1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-int64-1.0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_scalar[array_api_strictest-float64-1j-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestAsArrays.test_scalar_vs_scalar[array_api_strict] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestAsArrays.test_scalar_vs_scalar[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int8-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int16-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int32-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-int64-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint8-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint16-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint32-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-uint64-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float32-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-float64-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex64-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-complex128-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strict-bool-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int8-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int16-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int32-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-int64-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint8-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint16-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint32-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-uint64-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float32-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-float64-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex64-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-complex128-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-int8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-int16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-int32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-uint8] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-uint16] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-uint32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-uint64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-float32] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-complex64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-complex128] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAsArrays.test_array_vs_array[array_api_strictest-bool-bool] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____________ ERROR at setup of test_eager_shape[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestMetaNamespace.test_basic[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestMetaNamespace.test_basic[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestMetaNamespace.test_xp[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestMetaNamespace.test_xp[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestCapabilities.test_basic[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestCapabilities.test_basic[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestCapabilities.test_device[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestCapabilities.test_device[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestDiagIndices.test_basic[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_basic[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-1-2] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-1-4] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-1-7] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-2-2] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-2-4] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-2-7] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-3-2] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-3-4] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-3-7] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-4-2] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-4-4] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_ndim[array_api_strict-4-7] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-1-2] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-1-4] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-1-7] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-2-2] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-2-4] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-2-7] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-3-2] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-3-4] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-3-7] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-4-2] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-4-4] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_ndim[array_api_strictest-4-7] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestDiagIndices.test_empty[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_empty[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDiagIndices.test_validation[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestDiagIndices.test_validation[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestDiagIndices.test_device[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDiagIndices.test_device[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestTriIndices.test_basic[array_api_strict-tril] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestTriIndices.test_basic[array_api_strict-triu] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_basic[array_api_strictest-tril] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_basic[array_api_strictest-triu] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict--2-tril] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict--2-triu] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict--1-tril] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict--1-triu] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict-0-tril] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict-0-triu] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict-1-tril] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict-1-triu] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict-2-tril] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_offset[array_api_strict-2-triu] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestTriIndices.test_offset[array_api_strictest--2-tril] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestTriIndices.test_offset[array_api_strictest--2-triu] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestTriIndices.test_offset[array_api_strictest--1-tril] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestTriIndices.test_offset[array_api_strictest--1-triu] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_offset[array_api_strictest-0-tril] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_offset[array_api_strictest-0-triu] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_offset[array_api_strictest-1-tril] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_offset[array_api_strictest-1-triu] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_offset[array_api_strictest-2-tril] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_offset[array_api_strictest-2-triu] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_rectangular[array_api_strict-tril] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_rectangular[array_api_strict-triu] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestTriIndices.test_rectangular[array_api_strictest-tril] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestTriIndices.test_rectangular[array_api_strictest-triu] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_use_to_read[array_api_strict-tril] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_use_to_read[array_api_strict-triu] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestTriIndices.test_use_to_read[array_api_strictest-tril] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestTriIndices.test_use_to_read[array_api_strictest-triu] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_validation[array_api_strict-tril] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestTriIndices.test_validation[array_api_strict-triu] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestTriIndices.test_validation[array_api_strictest-tril] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestTriIndices.test_validation[array_api_strictest-triu] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestTriIndices.test_device[array_api_strict-tril] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestTriIndices.test_device[array_api_strict-triu] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_device[array_api_strictest-tril] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestTriIndices.test_device[array_api_strictest-triu] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestUnravelIndex.test_simple[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestUnravelIndex.test_simple[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestUnravelIndex.test_indices_scalar[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestUnravelIndex.test_indices_scalar[array_api_strictest] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestUnravelIndex.test_indices_2d[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestUnravelIndex.test_indices_2d[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestUnravelIndex.test_device[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestUnravelIndex.test_device[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestUnravelIndex.test_xp[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestUnravelIndex.test_xp[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestDefaultDType.test_basic[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDefaultDType.test_basic[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestDefaultDType.test_kind[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestDefaultDType.test_kind[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestDefaultDType.test_device[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestDefaultDType.test_device[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strict-int32-shape0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strict-int32-shape0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strict-int32-shape1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strict-int32-shape1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strict-float64-shape0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strict-float64-shape0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strict-float64-shape1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strict-float64-shape1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strictest-int32-shape0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strictest-int32-shape0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strictest-int32-shape1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strictest-int32-shape1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strictest-float64-shape0-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strictest-float64-shape0-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strictest-float64-shape1-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_simple[array_api_strictest-float64-shape1-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of test_lazy_apply_broadcast[array_api_strict-False] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of test_lazy_apply_broadcast[array_api_strict-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of test_lazy_apply_broadcast[array_api_strictest-False] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of test_lazy_apply_broadcast[array_api_strictest-True] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of test_lazy_apply_multi_output[array_api_strict-False] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of test_lazy_apply_multi_output[array_api_strict-True] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_lazy_apply_multi_output[array_api_strictest-False] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of test_lazy_apply_multi_output[array_api_strictest-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_multi_output_broadcast_dtype[array_api_strict-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_multi_output_broadcast_dtype[array_api_strict-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_multi_output_broadcast_dtype[array_api_strictest-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_apply_multi_output_broadcast_dtype[array_api_strictest-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of test_lazy_apply_none_shape_in_args[array_api_strict] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_lazy_apply_none_shape_in_args[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of test_lazy_apply_none_shape_broadcast[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of test_lazy_apply_device[array_api_strict-False] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of test_lazy_apply_device[array_api_strictest-False] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of test_lazy_apply_device[array_api_strictest-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of test_lazy_apply_arraylike[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of test_lazy_apply_arraylike[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of test_lazy_apply_scalars_and_nones[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of test_lazy_apply_scalars_and_nones[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of test_lazy_apply_kwargs[array_api_strict-False] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of test_lazy_apply_kwargs[array_api_strict-True] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of test_lazy_apply_kwargs[array_api_strictest-False] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of test_lazy_apply_kwargs[array_api_strictest-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_lazy_apply_raises[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of test_lazy_apply_raises[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestKron.test_basic[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestKron.test_basic[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestKron.test_kron_smoke[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestKron.test_kron_smoke[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strict-shape_a0-shape_b0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strict-shape_a1-shape_b1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strict-shape_a2-shape_b2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strict-shape_a3-shape_b3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strict-shape_a4-shape_b4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strict-shape_a5-shape_b5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strictest-shape_a0-shape_b0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strictest-shape_a1-shape_b1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strictest-shape_a2-shape_b2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strictest-shape_a3-shape_b3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strictest-shape_a4-shape_b4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestKron.test_kron_shape[array_api_strictest-shape_a5-shape_b5] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestKron.test_python_scalar[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestKron.test_python_scalar[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestKron.test_device[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestKron.test_device[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____________ ERROR at setup of TestKron.test_xp[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestKron.test_xp[array_api_strictest] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestAtLeastND.test_0D[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestAtLeastND.test_0D[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strict-input_shape0-0-expected_shape0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strict-input_shape1-1-expected_shape1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strict-input_shape2-2-expected_shape2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strict-input_shape3-3-expected_shape3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strict-input_shape4-5-expected_shape4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strictest-input_shape0-0-expected_shape0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strictest-input_shape1-1-expected_shape1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strictest-input_shape2-2-expected_shape2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strictest-input_shape3-3-expected_shape3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_1D_shapes[array_api_strictest-input_shape4-5-expected_shape4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestAtLeastND.test_1D_values[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestAtLeastND.test_1D_values[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strict-input_shape0-0-expected_shape0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strict-input_shape1-1-expected_shape1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strict-input_shape2-2-expected_shape2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strict-input_shape3-3-expected_shape3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strict-input_shape4-5-expected_shape4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strictest-input_shape0-0-expected_shape0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strictest-input_shape1-1-expected_shape1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strictest-input_shape2-2-expected_shape2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strictest-input_shape3-3-expected_shape3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_2D_shapes[array_api_strictest-input_shape4-5-expected_shape4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestAtLeastND.test_2D_values[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestAtLeastND.test_2D_values[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strict-input_shape0-0-expected_shape0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strict-input_shape1-1-expected_shape1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strict-input_shape2-2-expected_shape2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strict-input_shape3-3-expected_shape3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strict-input_shape4-5-expected_shape4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strictest-input_shape0-0-expected_shape0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strictest-input_shape1-1-expected_shape1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strictest-input_shape2-2-expected_shape2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strictest-input_shape3-3-expected_shape3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_3D_shapes[array_api_strictest-input_shape4-5-expected_shape4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestAtLeastND.test_3D_values[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestAtLeastND.test_3D_values[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strict-input_shape0-0-expected_shape0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strict-input_shape1-2-expected_shape1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strict-input_shape2-5-expected_shape2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strict-input_shape3-6-expected_shape3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strict-input_shape4-9-expected_shape4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strictest-input_shape0-0-expected_shape0] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strictest-input_shape1-2-expected_shape1] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strictest-input_shape2-5-expected_shape2] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strictest-input_shape3-6-expected_shape3] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAtLeastND.test_5D_shapes[array_api_strictest-input_shape4-9-expected_shape4] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestAtLeastND.test_5D_values[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestAtLeastND.test_5D_values[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestBroadcastShapes.test_xp[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestBroadcastShapes.test_xp[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestExpandDims.test_single_axis[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestExpandDims.test_single_axis[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestExpandDims.test_axis_tuple[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestExpandDims.test_axis_tuple[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestExpandDims.test_axis_out_of_range[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestExpandDims.test_axis_out_of_range[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestExpandDims.test_repeated_axis[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestExpandDims.test_repeated_axis[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestExpandDims.test_positive_negative_repeated[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestExpandDims.test_positive_negative_repeated[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestExpandDims.test_device[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestExpandDims.test_device[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestExpandDims.test_xp[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestExpandDims.test_xp[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestPad.test_simple[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestPad.test_simple[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestPad.test_fill_value[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestPad.test_fill_value[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestPad.test_ndim[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestPad.test_ndim[array_api_strictest] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestPad.test_mode_not_implemented[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestPad.test_mode_not_implemented[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestPad.test_device[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestPad.test_device[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____________ ERROR at setup of TestPad.test_xp[array_api_strict] ______________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestPad.test_xp[array_api_strictest] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestPad.test_tuple_width[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestPad.test_tuple_width[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestPad.test_sequence_of_tuples_width[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestPad.test_sequence_of_tuples_width[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestSearchsorted.test_input_validation[array_api_strict] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_input_validation[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-False-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-False-True-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-False-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted0-True-True-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-False-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-False-True-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-False-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strict-searchsorted1-True-True-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-False-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-False-True-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-False-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted0-True-True-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-False-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-False-True-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-False-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-0-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-0-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-0-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-0-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-1-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-1-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-1-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-1-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-2-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-2-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-2-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-2-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-11-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-11-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-11-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-11-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-1000-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-1000-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-1000-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-1000-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-10001-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-10001-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-10001-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-10001-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape7-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape7-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape7-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape7-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape8-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape8-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape8-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape8-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape9-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape9-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape9-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape9-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape10-False-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape10-False-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape10-True-left] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSearchsorted.test_nd[array_api_strictest-searchsorted1-True-True-shape10-True-right] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNUnique.test_simple[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNUnique.test_simple[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNUnique.test_empty[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNUnique.test_empty[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNUnique.test_size1[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNUnique.test_size1[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNUnique.test_all_equal[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNUnique.test_all_equal[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestNUnique.test_nan[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNUnique.test_nan[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNUnique.test_device[array_api_strict-0] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNUnique.test_device[array_api_strict-1] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNUnique.test_device[array_api_strict-2] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNUnique.test_device[array_api_strictest-0] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNUnique.test_device[array_api_strictest-1] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNUnique.test_device[array_api_strictest-2] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestNUnique.test_xp[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNUnique.test_xp[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestSetDiff1D.test_setdiff1d[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestSetDiff1D.test_assume_unique[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape20-shape10-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape20-shape10-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape20-shape11-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape20-shape11-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape20-shape12-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape20-shape12-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape21-shape10-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape21-shape10-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape21-shape11-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape21-shape11-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape21-shape12-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape21-shape12-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape22-shape10-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape22-shape10-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape22-shape11-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape22-shape11-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape22-shape12-True] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestSetDiff1D.test_shapes[array_api_strict-shape22-shape12-False] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestSetDiff1D.test_python_scalar[array_api_strict-True] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestSetDiff1D.test_python_scalar[array_api_strict-False] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestSetDiff1D.test_device[array_api_strict-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestSetDiff1D.test_device[array_api_strict-False] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestSetDiff1D.test_xp[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestIsIn.test_simple[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestIsIn.test_simple[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestIsIn.test_device[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestIsIn.test_device[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestIsIn.test_assume_unique_and_invert[array_api_strict] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestIsIn.test_assume_unique_and_invert[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestIsIn.test_kind[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestIsIn.test_kind[array_api_strictest] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestUnion1d.test_simple[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestUnion1d.test_2d[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestUnion1d.test_3d[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestUnion1d.test_device[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestPartition.test_1d[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestPartition.test_1d[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestPartition.test_nd[array_api_strict-2] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestPartition.test_nd[array_api_strict-3] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestPartition.test_nd[array_api_strict-4] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestPartition.test_nd[array_api_strictest-2] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestPartition.test_nd[array_api_strictest-3] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestPartition.test_nd[array_api_strictest-4] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestPartition.test_input_validation[array_api_strict] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestPartition.test_input_validation[array_api_strictest] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestArgpartition.test_1d[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestArgpartition.test_1d[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestArgpartition.test_nd[array_api_strict-2] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestArgpartition.test_nd[array_api_strict-3] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestArgpartition.test_nd[array_api_strict-4] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestArgpartition.test_nd[array_api_strictest-2] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestArgpartition.test_nd[array_api_strictest-3] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestArgpartition.test_nd[array_api_strictest-4] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestArgpartition.test_input_validation[array_api_strict] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestArgpartition.test_input_validation[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestCov.test_basic[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestCov.test_basic[array_api_strictest] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestCov.test_complex[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestCov.test_complex[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCov.test_complex_with_weights[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCov.test_complex_with_weights[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestCov.test_empty[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestCov.test_empty[array_api_strictest] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestCov.test_combination[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestCov.test_combination[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestCov.test_device[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestCov.test_device[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____________ ERROR at setup of TestCov.test_xp[array_api_strict] ______________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestCov.test_xp[array_api_strictest] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestCov.test_batch[array_api_strict] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestCov.test_batch[array_api_strictest] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestCov.test_bias[array_api_strict-True] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestCov.test_bias[array_api_strict-False] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestCov.test_bias[array_api_strict-0] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestCov.test_bias[array_api_strict-1] ____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestCov.test_bias[array_api_strictest-True] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestCov.test_bias[array_api_strictest-False] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestCov.test_bias[array_api_strictest-0] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestCov.test_bias[array_api_strictest-1] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestCov.test_bias_batch[array_api_strict-True] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestCov.test_bias_batch[array_api_strict-False] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestCov.test_bias_batch[array_api_strict-0] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestCov.test_bias_batch[array_api_strict-1] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestCov.test_bias_batch[array_api_strictest-True] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestCov.test_bias_batch[array_api_strictest-False] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestCov.test_bias_batch[array_api_strictest-0] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestCov.test_bias_batch[array_api_strictest-1] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestCov.test_correction[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestCov.test_correction[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestCov.test_correction_float[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestCov.test_correction_float[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____________ ERROR at setup of TestCov.test_axis[array_api_strict] _____________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestCov.test_axis[array_api_strictest] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestCov.test_frequency_weights[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCov.test_frequency_weights[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of TestCov.test_weights[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestCov.test_weights[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestCov.test_both_weights[array_api_strict] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestCov.test_both_weights[array_api_strictest] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestCov.test_batch_with_weights[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCov.test_batch_with_weights[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestCov.test_axis_with_weights[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCov.test_axis_with_weights[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestCov.test_axis_out_of_bounds[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCov.test_axis_out_of_bounds[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestCov.test_weights_wrong_ndim[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCov.test_weights_wrong_ndim[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestCov.test_weights_wrong_length[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestCov.test_weights_wrong_length[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanMin.test_simple[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMin.test_simple[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanMin.test_bigger[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMin.test_bigger[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNanMin.test_with_infinity[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanMin.test_with_infinity[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanMin.test_scalar[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMin.test_scalar[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanMin.test_all_nan_slice_2d[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanMin.test_all_nan_slice_2d[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMin.test_device[array_api_strict-None] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanMin.test_device[array_api_strict-0] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanMin.test_device[array_api_strict-1] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNanMin.test_device[array_api_strictest-None] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMin.test_device[array_api_strictest-0] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMin.test_device[array_api_strictest-1] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanMin.test_xp[array_api_strict-0-expected_list0] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanMin.test_xp[array_api_strict-1-expected_list1] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestNanMin.test_xp[array_api_strictest-0-expected_list0] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestNanMin.test_xp[array_api_strictest-1-expected_list1] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanMax.test_simple[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMax.test_simple[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanMax.test_bigger[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMax.test_bigger[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNanMax.test_with_infinity[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanMax.test_with_infinity[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanMax.test_scalar[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMax.test_scalar[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanMax.test_all_nan_slice_2d[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanMax.test_all_nan_slice_2d[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMax.test_device[array_api_strict-None] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanMax.test_device[array_api_strict-0] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanMax.test_device[array_api_strict-1] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNanMax.test_device[array_api_strictest-None] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMax.test_device[array_api_strictest-0] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMax.test_device[array_api_strictest-1] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanMax.test_xp[array_api_strict-0-expected_list0] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanMax.test_xp[array_api_strict-1-expected_list1] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestNanMax.test_xp[array_api_strictest-0-expected_list0] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestNanMax.test_xp[array_api_strictest-1-expected_list1] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanSum.test_simple[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanSum.test_simple[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanSum.test_bigger[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanSum.test_bigger[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNanSum.test_all_nan_slice[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanSum.test_all_nan_slice[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of TestNanSum.test_scalar[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanSum.test_scalar[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanSum.test_device[array_api_strict-None] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanSum.test_device[array_api_strict-0] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanSum.test_device[array_api_strict-1] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNanSum.test_device[array_api_strictest-None] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanSum.test_device[array_api_strictest-0] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanSum.test_device[array_api_strictest-1] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanSum.test_xp[array_api_strict-0-expected_list0] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanSum.test_xp[array_api_strict-1-expected_list1] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestNanSum.test_xp[array_api_strictest-0-expected_list0] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of TestNanSum.test_xp[array_api_strictest-1-expected_list1] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanMean.test_simple[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMean.test_simple[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanMean.test_bigger[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMean.test_bigger[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestNanMean.test_all_nan_slice[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of TestNanMean.test_all_nan_slice[array_api_strictest] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of TestNanMean.test_scalar[array_api_strict] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMean.test_scalar[array_api_strictest] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMean.test_device[array_api_strict-None] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMean.test_device[array_api_strict-0] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ________ ERROR at setup of TestNanMean.test_device[array_api_strict-1] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestNanMean.test_device[array_api_strictest-None] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMean.test_device[array_api_strictest-0] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestNanMean.test_device[array_api_strictest-1] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanMean.test_xp[array_api_strict-0-expected_list0] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of TestNanMean.test_xp[array_api_strict-1-expected_list1] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanMean.test_xp[array_api_strictest-0-expected_list0] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestNanMean.test_xp[array_api_strictest-1-expected_list1] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestAsNumPyArray.test_basic[array_api_strict] ________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ______ ERROR at setup of TestAsNumPyArray.test_basic[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _______ ERROR at setup of TestAsNumPyArray.test_device[array_api_strict] _______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of TestAsNumPyArray.test_device[array_api_strictest] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_equal_close_basic[array_api_strict-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_equal_close_basic[array_api_strict-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_equal_close_basic[array_api_strictest-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_equal_close_basic[array_api_strictest-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_shape_dtype[array_api_strict-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_shape_dtype[array_api_strict-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_shape_dtype[array_api_strict-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_shape_dtype[array_api_strict-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_shape_dtype[array_api_strictest-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_shape_dtype[array_api_strictest-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_shape_dtype[array_api_strictest-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_shape_dtype[array_api_strictest-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_namespace[array_api_strict-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_namespace[array_api_strict-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_namespace[array_api_strict-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_namespace[array_api_strict-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_namespace[array_api_strictest-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_namespace[array_api_strictest-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_namespace[array_api_strictest-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_namespace[array_api_strictest-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_shape[array_api_strict-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_shape[array_api_strict-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_shape[array_api_strict-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_shape[array_api_strict-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_shape[array_api_strictest-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_shape[array_api_strictest-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_shape[array_api_strictest-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_shape[array_api_strictest-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_dtype[array_api_strict-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_dtype[array_api_strict-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_dtype[array_api_strict-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_dtype[array_api_strict-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_dtype[array_api_strictest-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_dtype[array_api_strictest-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_dtype[array_api_strictest-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_dtype[array_api_strictest-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_scalar[array_api_strict-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_scalar[array_api_strict-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_scalar[array_api_strict-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_scalar[array_api_strict-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_scalar[array_api_strictest-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_scalar[array_api_strictest-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_scalar[array_api_strictest-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_check_scalar[array_api_strictest-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_close_tolerance[array_api_strict-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_close_tolerance[array_api_strict-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_close_tolerance[array_api_strictest-int64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_close_tolerance[array_api_strictest-float64] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_less[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_less[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_none_shape[array_api_strict-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_none_shape[array_api_strict-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_none_shape[array_api_strict-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_none_shape[array_api_strict-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_device[array_api_strict-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_device[array_api_strict-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_device[array_api_strict-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_device[array_api_strict-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_device[array_api_strictest-assert_equal] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_device[array_api_strictest-assert_close] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_device[array_api_strictest-assert_less] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_device[array_api_strictest-assert_close_nulp] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_close_nulp[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestAssertEqualCloseLess.test_assert_close_nulp[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __________ ERROR at setup of test_lazy_xp_function[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of test_lazy_xp_function[array_api_strictest] _________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of test_lazy_xp_function_static_params[array_api_strict] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_lazy_xp_function_static_params[array_api_strictest] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of test_lazy_xp_function_cython_ufuncs[array_api_strict] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_lazy_xp_function_cython_ufuncs[array_api_strictest] __ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_lazy_xp_function_classes[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_lazy_xp_function_classes[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_static_methods_preserved[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_static_methods_preserved[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_static_methods_wrapped[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_static_methods_wrapped[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_class_methods_wrapped[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_class_methods_wrapped[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_circumvention[array_api_strict] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of TestLazyXpFunctionClasses.test_circumvention[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError __ ERROR at setup of test_lazy_xp_function_opaque_wrappers[array_api_strict] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _ ERROR at setup of test_lazy_xp_function_opaque_wrappers[array_api_strictest] _ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _____ ERROR at setup of test_lazy_xp_function_recursive[array_api_strict] ______ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of test_lazy_xp_function_recursive[array_api_strictest] ____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___________ ERROR at setup of test_lazy_xp_modules[array_api_strict] ___________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError _________ ERROR at setup of test_lazy_xp_modules[array_api_strictest] __________ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ____ ERROR at setup of test_patch_lazy_xp_functions_iter[array_api_strict] _____ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError ___ ERROR at setup of test_patch_lazy_xp_functions_iter[array_api_strictest] ___ library = request = > @pytest.fixture def xp( library: Backend, request: pytest.FixtureRequest ) -> Generator[ArrayNamespace]: # numpydoc ignore=PR01,RT03 """ Parameterized fixture that iterates on all libraries. Returns ------- The current array namespace. """ if library == Backend.NUMPY_READONLY: yield NumPyReadOnly() # type: ignore[misc] # pyright: ignore[reportReturnType] return xp = pytest.importorskip(library.modname) # Possibly wrap module with array_api_compat xp = array_namespace(xp.empty(0)) if library.like(Backend.ARRAY_API_STRICT): > xp.set_array_api_strict_flags(api_version="2025.12") tests/main/conftest.py:134: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ def set_array_api_strict_flags( *, api_version=None, boolean_indexing=None, data_dependent_shapes=None, enabled_extensions=None, ): """ Set the array-api-strict flags to the specified values. Flags are global variables that enable or disable array-api-strict behaviors. The flags can also be changed by setting :ref:`environment variables `. .. note:: This function is **not** part of the array API standard. It only exists in array-api-strict. Parameters ---------- api_version : str, optional The version of the standard to use. Supported versions are: ``{supported_versions}``. The default version number is ``{default_version!r}``. Note that 2021.12 is supported, but currently gives the same thing as 2022.12 (except that the fft extension will be disabled). 2023.12 support is experimental. Some features in 2023.12 may still be missing, and it hasn't been fully tested. A future version of array-api-strict will change the default version to 2023.12. boolean_indexing : bool, optional Whether indexing by a boolean array is supported. This flag is enabled by default. Note that although boolean array indexing does result in data-dependent shapes, this flag is independent of the `data_dependent_shapes` flag (see below). data_dependent_shapes : bool, optional Whether data-dependent shapes are enabled in array-api-strict. This flag is enabled by default. Array libraries that use computation graphs may not be able to support functions whose output shapes depend on the input data. The functions that make use of data-dependent shapes, and are therefore disabled by setting this flag to False are - `unique_all()`, `unique_counts()`, `unique_inverse()`, and `unique_values()`. - `nonzero()` - `repeat()` when the `repeats` argument is an array (requires the 2023.12 version of the standard) Note that while boolean indexing is also data-dependent, it is controlled by a separate `boolean_indexing` flag (see above). See https://data-apis.org/array-api/latest/design_topics/data_dependent_output_shapes.html for more details. enabled_extensions : list of str, optional A list of extensions that are enabled in array-api-strict. The default is ``{default_extensions}``. Note that some extensions require a minimum version of the standard. Examples -------- >>> from array_api_strict import set_array_api_strict_flags >>> # Set the standard version to 2021.12 >>> set_array_api_strict_flags(api_version="2021.12") >>> # Disable data-dependent shapes and boolean indexing >>> set_array_api_strict_flags(data_dependent_shapes=False, boolean_indexing=False) >>> # Enable only the linalg extension (disable the fft extension) >>> set_array_api_strict_flags(enabled_extensions=["linalg"]) See Also -------- get_array_api_strict_flags: Get the current values of flags. reset_array_api_strict_flags: Reset the flags to their default values. ArrayAPIStrictFlags: A context manager to temporarily set the flags. """ global API_VERSION, BOOLEAN_INDEXING, DATA_DEPENDENT_SHAPES, ENABLED_EXTENSIONS if api_version is not None: if api_version not in supported_versions: > raise ValueError(f"Unsupported standard version {api_version!r}") E ValueError: Unsupported standard version '2025.12' /usr/lib/python3.14/site-packages/array_api_strict/_flags.py:142: ValueError =========================== short test summary info ============================ SKIPPED [144] tests/main/test_at.py:80: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [144] tests/main/test_at.py:80: could not import 'mparray': No module named 'mparray' SKIPPED [144] tests/main/test_at.py:80: could not import 'cupy': No module named 'cupy' SKIPPED [288] tests/main/test_at.py:80: could not import 'torch': No module named 'torch' SKIPPED [144] tests/main/test_at.py:80: could not import 'dask.array': No module named 'dask' SKIPPED [144] tests/main/test_at.py:80: Backend.SPARSE: read-only backend without .at support SKIPPED [288] tests/main/test_at.py:80: could not import 'jax.numpy': No module named 'jax' SKIPPED [8] tests/main/test_at.py:155: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [8] tests/main/test_at.py:155: could not import 'mparray': No module named 'mparray' SKIPPED [8] tests/main/test_at.py:155: could not import 'cupy': No module named 'cupy' SKIPPED [16] tests/main/test_at.py:155: could not import 'torch': No module named 'torch' SKIPPED [8] tests/main/test_at.py:155: could not import 'dask.array': No module named 'dask' SKIPPED [8] tests/main/test_at.py:155: Backend.SPARSE: read-only backend without .at support SKIPPED [16] tests/main/test_at.py:155: could not import 'jax.numpy': No module named 'jax' SKIPPED [32] tests/main/test_at.py:211: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [32] tests/main/test_at.py:211: could not import 'mparray': No module named 'mparray' SKIPPED [32] tests/main/test_at.py:211: could not import 'cupy': No module named 'cupy' SKIPPED [64] tests/main/test_at.py:211: could not import 'torch': No module named 'torch' SKIPPED [32] tests/main/test_at.py:211: could not import 'dask.array': No module named 'dask' SKIPPED [32] tests/main/test_at.py:211: Backend.SPARSE: read-only backend without .at support SKIPPED [64] tests/main/test_at.py:211: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_at.py:272: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [1] tests/main/test_at.py:272: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_at.py:272: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_at.py:272: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_at.py:272: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_at.py:272: Backend.SPARSE: read-only backend without .at support SKIPPED [2] tests/main/test_at.py:272: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_at.py:279: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [2] tests/main/test_at.py:279: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_at.py:279: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_at.py:279: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_at.py:279: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_at.py:279: Backend.SPARSE: read-only backend without .at support SKIPPED [4] tests/main/test_at.py:279: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_at.py:288: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [2] tests/main/test_at.py:288: Backend.NUMPY: np.generic is read-only SKIPPED [2] tests/main/test_at.py:288: Backend.NUMPY_READONLY: read-only backend SKIPPED [4] tests/main/test_at.py:288: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_at.py:288: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_at.py:288: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_at.py:288: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_at.py:288: Backend.SPARSE: read-only backend without .at support SKIPPED [2] tests/main/test_at.py:288: Backend.SPARSE: read-only backend SKIPPED [4] tests/main/test_at.py:288: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_at.py:288: Backend.JAX: read-only backend SKIPPED [2] tests/main/test_at.py:288: Backend.JAX_GPU: read-only backend SKIPPED [1] tests/main/test_at.py:334: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [1] tests/main/test_at.py:334: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_at.py:334: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_at.py:334: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_at.py:334: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_at.py:334: Backend.SPARSE: read-only backend without .at support SKIPPED [2] tests/main/test_at.py:334: could not import 'jax.numpy': No module named 'jax' SKIPPED [12] tests/main/test_creation.py:17: could not import 'mparray': No module named 'mparray' SKIPPED [12] tests/main/test_creation.py:17: could not import 'cupy': No module named 'cupy' SKIPPED [24] tests/main/test_creation.py:17: could not import 'torch': No module named 'torch' SKIPPED [12] tests/main/test_creation.py:17: could not import 'dask.array': No module named 'dask' SKIPPED [12] tests/main/test_creation.py:17: could not import 'sparse': No module named 'sparse' SKIPPED [24] tests/main/test_creation.py:17: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:30: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:30: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:30: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:30: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:30: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_creation.py:30: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:39: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:39: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:39: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:39: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:39: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_creation.py:39: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:49: Backend.ARRAY_API_STRICTEST: backend doesn't support Boolean indexing SKIPPED [1] tests/main/test_creation.py:49: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:49: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:49: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:49: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:49: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_creation.py:49: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:59: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:59: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_creation.py:59: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:59: Backend.TORCH_GPU: Puts Pytorch into a bad state. SKIPPED [1] tests/main/test_creation.py:59: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:59: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_creation.py:59: could not import 'jax.numpy': No module named 'jax' SKIPPED [8] tests/main/test_creation.py:71: could not import 'mparray': No module named 'mparray' SKIPPED [8] tests/main/test_creation.py:71: could not import 'cupy': No module named 'cupy' SKIPPED [16] tests/main/test_creation.py:71: could not import 'torch': No module named 'torch' SKIPPED [8] tests/main/test_creation.py:71: could not import 'dask.array': No module named 'dask' SKIPPED [8] tests/main/test_creation.py:71: could not import 'sparse': No module named 'sparse' SKIPPED [16] tests/main/test_creation.py:71: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:82: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:82: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:82: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:82: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:82: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_creation.py:82: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:89: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:89: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:89: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:89: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:89: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_creation.py:89: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:97: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:97: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:97: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:97: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:97: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_creation.py:97: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:101: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:101: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:101: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:101: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:101: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_creation.py:101: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:111: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:111: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:111: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:111: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:111: Backend.SPARSE: read-only backend without .at support SKIPPED [2] tests/main/test_creation.py:111: could not import 'jax.numpy': No module named 'jax' SKIPPED [81] tests/main/test_creation.py:126: could not import 'mparray': No module named 'mparray' SKIPPED [81] tests/main/test_creation.py:126: could not import 'cupy': No module named 'cupy' SKIPPED [162] tests/main/test_creation.py:126: could not import 'torch': No module named 'torch' SKIPPED [81] tests/main/test_creation.py:126: could not import 'dask.array': No module named 'dask' SKIPPED [81] tests/main/test_creation.py:126: Backend.SPARSE: read-only backend without .at support SKIPPED [162] tests/main/test_creation.py:126: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:137: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:137: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:137: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:137: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:137: Backend.SPARSE: read-only backend without .at support SKIPPED [2] tests/main/test_creation.py:137: could not import 'jax.numpy': No module named 'jax' SKIPPED [10] tests/main/test_creation.py:141: could not import 'mparray': No module named 'mparray' SKIPPED [10] tests/main/test_creation.py:141: could not import 'cupy': No module named 'cupy' SKIPPED [20] tests/main/test_creation.py:141: could not import 'torch': No module named 'torch' SKIPPED [10] tests/main/test_creation.py:141: could not import 'dask.array': No module named 'dask' SKIPPED [10] tests/main/test_creation.py:141: Backend.SPARSE: read-only backend without .at support SKIPPED [20] tests/main/test_creation.py:141: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:167: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:167: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:167: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:167: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:167: Backend.SPARSE: read-only backend without .at support SKIPPED [2] tests/main/test_creation.py:167: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_creation.py:171: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_creation.py:171: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_creation.py:171: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_creation.py:171: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_creation.py:171: Backend.SPARSE: read-only backend without .at support SKIPPED [2] tests/main/test_creation.py:171: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_deprecation.py:8: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_deprecation.py:8: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_deprecation.py:8: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_deprecation.py:8: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_deprecation.py:8: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_deprecation.py:8: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_deprecation.py:12: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_deprecation.py:12: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_deprecation.py:12: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_deprecation.py:12: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_deprecation.py:12: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_deprecation.py:12: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:44: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:44: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:44: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:44: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:44: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:44: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:51: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:51: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:51: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:51: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:51: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:51: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:61: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:61: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:61: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:61: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:61: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:61: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:69: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:69: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:69: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:69: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:69: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:69: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:92: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:92: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:92: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:92: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:92: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:92: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_elementwise.py:109: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_elementwise.py:109: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_elementwise.py:109: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_elementwise.py:109: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_elementwise.py:109: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_elementwise.py:109: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:127: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:127: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:127: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:127: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:127: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:127: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:134: Backend.ARRAY_API_STRICTEST: no boolean indexing -> run everywhere SKIPPED [1] tests/main/test_elementwise.py:134: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:134: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:134: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:134: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:134: Backend.SPARSE: no indexing by sparse array -> run everywhere SKIPPED [2] tests/main/test_elementwise.py:134: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:154: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:154: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:154: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:154: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:154: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:154: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:164: Backend.NUMPY_READONLY: xp=xp SKIPPED [1] tests/main/test_elementwise.py:164: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:164: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:164: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:164: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:164: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:164: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:172: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:172: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:172: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:172: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:172: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:172: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:181: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:181: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:181: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:181: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:181: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:181: could not import 'jax.numpy': No module named 'jax' SKIPPED [46] tests/main/test_elementwise.py:284: could not import 'mparray': No module named 'mparray' SKIPPED [46] tests/main/test_elementwise.py:284: could not import 'cupy': No module named 'cupy' SKIPPED [92] tests/main/test_elementwise.py:284: could not import 'torch': No module named 'torch' SKIPPED [46] tests/main/test_elementwise.py:284: could not import 'dask.array': No module named 'dask' SKIPPED [46] tests/main/test_elementwise.py:284: could not import 'sparse': No module named 'sparse' SKIPPED [92] tests/main/test_elementwise.py:284: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_elementwise.py:330: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_elementwise.py:330: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_elementwise.py:330: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_elementwise.py:330: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_elementwise.py:330: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_elementwise.py:330: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:342: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:342: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:342: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:342: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:342: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:342: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:348: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:348: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:348: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:348: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:348: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:348: could not import 'jax.numpy': No module named 'jax' SKIPPED [3] tests/main/test_elementwise.py:354: could not import 'mparray': No module named 'mparray' SKIPPED [3] tests/main/test_elementwise.py:354: could not import 'cupy': No module named 'cupy' SKIPPED [6] tests/main/test_elementwise.py:354: could not import 'torch': No module named 'torch' SKIPPED [3] tests/main/test_elementwise.py:354: could not import 'dask.array': No module named 'dask' SKIPPED [3] tests/main/test_elementwise.py:354: could not import 'sparse': No module named 'sparse' SKIPPED [6] tests/main/test_elementwise.py:354: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_elementwise.py:367: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_elementwise.py:367: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_elementwise.py:367: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_elementwise.py:367: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_elementwise.py:367: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_elementwise.py:367: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:374: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:374: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:374: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:374: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:374: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:374: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:383: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:383: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:383: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:383: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:383: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:383: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:400: Backend.ARRAY_API_STRICTEST: unknown shape SKIPPED [1] tests/main/test_elementwise.py:400: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:400: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:400: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:400: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:400: Backend.SPARSE: index by sparse array SKIPPED [2] tests/main/test_elementwise.py:400: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:409: Backend.ARRAY_API_STRICTEST: unknown shape SKIPPED [1] tests/main/test_elementwise.py:409: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:409: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:409: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:409: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:409: Backend.SPARSE: index by sparse array SKIPPED [2] tests/main/test_elementwise.py:409: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:418: Backend.NUMPY_READONLY: xp=xp SKIPPED [1] tests/main/test_elementwise.py:418: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:418: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:418: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:418: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:418: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:418: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:435: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:435: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:435: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:435: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:435: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:435: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_elementwise.py:440: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_elementwise.py:440: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_elementwise.py:440: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_elementwise.py:440: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_elementwise.py:440: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_elementwise.py:440: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:447: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:447: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:447: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:447: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:447: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:447: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:462: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:462: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:462: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:462: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:462: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:462: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:466: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:466: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:466: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:466: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:466: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:466: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:470: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:470: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:470: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:470: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:470: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:470: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:474: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:474: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:474: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:474: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:474: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:474: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:478: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:478: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:478: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:478: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:478: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:478: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:493: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:493: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:493: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:493: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:493: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:493: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:506: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:506: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:506: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:506: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:506: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:506: could not import 'jax.numpy': No module named 'jax' SKIPPED [5] tests/main/test_elementwise.py:511: could not import 'mparray': No module named 'mparray' SKIPPED [5] tests/main/test_elementwise.py:511: could not import 'cupy': No module named 'cupy' SKIPPED [10] tests/main/test_elementwise.py:511: could not import 'torch': No module named 'torch' SKIPPED [5] tests/main/test_elementwise.py:511: could not import 'dask.array': No module named 'dask' SKIPPED [5] tests/main/test_elementwise.py:511: could not import 'sparse': No module named 'sparse' SKIPPED [10] tests/main/test_elementwise.py:511: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:576: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:576: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:576: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:576: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:576: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:576: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:596: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:596: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:596: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:596: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:596: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:596: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_elementwise.py:603: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_elementwise.py:603: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_elementwise.py:603: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_elementwise.py:603: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_elementwise.py:603: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_elementwise.py:603: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:608: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:608: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:608: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:608: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:608: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:608: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:616: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:616: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:616: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:616: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:616: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:616: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:620: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:620: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:620: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:620: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:620: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:620: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:625: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:625: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:625: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:625: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:625: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:625: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:631: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:631: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:631: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:631: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:631: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:631: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:666: Backend.MPARRAY: negative zero not supported SKIPPED [1] tests/main/test_elementwise.py:666: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:666: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:666: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:666: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:666: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:672: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:672: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:672: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:672: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:672: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:672: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:678: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:678: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:678: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:678: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:678: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:678: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:686: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:686: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:686: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:686: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:686: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:686: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:694: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:694: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_elementwise.py:694: Backend.TORCH: materialize 'meta' device SKIPPED [1] tests/main/test_elementwise.py:694: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:694: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:694: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:694: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:701: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:701: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:701: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:701: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:701: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:701: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_elementwise.py:706: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_elementwise.py:706: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_elementwise.py:706: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_elementwise.py:706: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_elementwise.py:706: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_elementwise.py:706: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:716: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:716: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:716: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:716: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:716: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:716: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:723: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:723: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:723: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:723: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:723: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:723: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:728: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:728: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:728: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:728: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:728: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:728: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:734: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:734: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:734: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:734: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:734: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:734: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_elementwise.py:739: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_elementwise.py:739: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_elementwise.py:739: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_elementwise.py:739: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_elementwise.py:739: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_elementwise.py:739: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:750: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:750: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:750: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:750: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:750: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:750: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:757: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:757: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:757: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:757: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:757: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:757: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_elementwise.py:762: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_elementwise.py:762: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_elementwise.py:762: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_elementwise.py:762: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_elementwise.py:762: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_elementwise.py:762: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_helpers.py:36: Backend.ARRAY_API_STRICTEST: no unique_inverse SKIPPED [2] tests/main/test_helpers.py:36: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_helpers.py:36: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_helpers.py:36: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_helpers.py:36: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_helpers.py:36: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_helpers.py:36: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:56: Backend.ARRAY_API_STRICTEST: no unique_inverse SKIPPED [1] tests/main/test_helpers.py:56: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_helpers.py:56: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_helpers.py:56: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_helpers.py:56: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:56: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_helpers.py:56: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:61: Backend.ARRAY_API_STRICTEST: no unique_inverse SKIPPED [1] tests/main/test_helpers.py:61: Backend.NUMPY_READONLY: xp=xp SKIPPED [1] tests/main/test_helpers.py:61: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_helpers.py:61: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_helpers.py:61: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_helpers.py:61: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:61: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_helpers.py:61: could not import 'jax.numpy': No module named 'jax' SKIPPED [16] tests/main/test_helpers.py:71: could not import 'mparray': No module named 'mparray' SKIPPED [16] tests/main/test_helpers.py:71: could not import 'cupy': No module named 'cupy' SKIPPED [32] tests/main/test_helpers.py:71: could not import 'torch': No module named 'torch' SKIPPED [16] tests/main/test_helpers.py:71: could not import 'dask.array': No module named 'dask' SKIPPED [16] tests/main/test_helpers.py:71: could not import 'sparse': No module named 'sparse' SKIPPED [32] tests/main/test_helpers.py:71: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:114: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_helpers.py:114: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_helpers.py:114: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_helpers.py:114: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:114: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_helpers.py:114: could not import 'jax.numpy': No module named 'jax' SKIPPED [169] tests/main/test_helpers.py:135: could not import 'mparray': No module named 'mparray' SKIPPED [169] tests/main/test_helpers.py:135: could not import 'cupy': No module named 'cupy' SKIPPED [338] tests/main/test_helpers.py:135: could not import 'torch': No module named 'torch' SKIPPED [169] tests/main/test_helpers.py:135: could not import 'dask.array': No module named 'dask' SKIPPED [169] tests/main/test_helpers.py:135: could not import 'sparse': No module named 'sparse' SKIPPED [338] tests/main/test_helpers.py:135: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:167: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [1] tests/main/test_helpers.py:167: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_helpers.py:167: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_helpers.py:167: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_helpers.py:167: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:167: Backend.SPARSE: index by sparse array SKIPPED [2] tests/main/test_helpers.py:167: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:189: Backend.NUMPY_READONLY: namespace tests SKIPPED [1] tests/main/test_helpers.py:189: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_helpers.py:189: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_helpers.py:189: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_helpers.py:189: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:189: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_helpers.py:189: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:195: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:201: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_helpers.py:201: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_helpers.py:201: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_helpers.py:201: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:201: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_helpers.py:201: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:207: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_helpers.py:207: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_helpers.py:207: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_helpers.py:207: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:207: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_helpers.py:207: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:213: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_helpers.py:213: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_helpers.py:213: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_helpers.py:213: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_helpers.py:213: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_helpers.py:213: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_helpers.py:349: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_helpers.py:366: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_helpers.py:375: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_helpers.py:399: could not import 'jax': No module named 'jax' SKIPPED [2] tests/main/test_helpers.py:411: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:19: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:19: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:19: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:19: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:19: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:19: could not import 'jax.numpy': No module named 'jax' SKIPPED [12] tests/main/test_indexing.py:25: could not import 'mparray': No module named 'mparray' SKIPPED [12] tests/main/test_indexing.py:25: could not import 'cupy': No module named 'cupy' SKIPPED [24] tests/main/test_indexing.py:25: could not import 'torch': No module named 'torch' SKIPPED [12] tests/main/test_indexing.py:25: could not import 'dask.array': No module named 'dask' SKIPPED [12] tests/main/test_indexing.py:25: could not import 'sparse': No module named 'sparse' SKIPPED [24] tests/main/test_indexing.py:25: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:34: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:34: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:34: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:34: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:34: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:34: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:39: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:39: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:39: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:39: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:39: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:39: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:45: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:45: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:45: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:45: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:45: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:45: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_indexing.py:67: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_indexing.py:67: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_indexing.py:67: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_indexing.py:67: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_indexing.py:67: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_indexing.py:67: could not import 'jax.numpy': No module named 'jax' SKIPPED [10] tests/main/test_indexing.py:78: could not import 'mparray': No module named 'mparray' SKIPPED [10] tests/main/test_indexing.py:78: could not import 'cupy': No module named 'cupy' SKIPPED [20] tests/main/test_indexing.py:78: could not import 'torch': No module named 'torch' SKIPPED [10] tests/main/test_indexing.py:78: could not import 'dask.array': No module named 'dask' SKIPPED [10] tests/main/test_indexing.py:78: could not import 'sparse': No module named 'sparse' SKIPPED [20] tests/main/test_indexing.py:78: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_indexing.py:91: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_indexing.py:91: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_indexing.py:91: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_indexing.py:91: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_indexing.py:91: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_indexing.py:91: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_indexing.py:102: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_indexing.py:102: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_indexing.py:102: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_indexing.py:102: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_indexing.py:102: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_indexing.py:102: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_indexing.py:117: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_indexing.py:117: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_indexing.py:117: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_indexing.py:117: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_indexing.py:117: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_indexing.py:117: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_indexing.py:128: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_indexing.py:128: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_indexing.py:128: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_indexing.py:128: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_indexing.py:128: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_indexing.py:128: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:145: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:145: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:145: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:145: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:145: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:145: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:163: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:163: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:163: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:163: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:163: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:163: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:171: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:171: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:171: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:171: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:171: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:171: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:184: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:184: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:184: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:184: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:184: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:184: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_indexing.py:191: Backend.NUMPY_READONLY: xp=xp SKIPPED [1] tests/main/test_indexing.py:191: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_indexing.py:191: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_indexing.py:191: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_indexing.py:191: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_indexing.py:191: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_indexing.py:191: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_inspection.py:11: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_inspection.py:11: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_inspection.py:11: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_inspection.py:11: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_inspection.py:11: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_inspection.py:11: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_inspection.py:14: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_inspection.py:14: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_inspection.py:14: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_inspection.py:14: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_inspection.py:14: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_inspection.py:14: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_inspection.py:23: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_inspection.py:23: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_inspection.py:23: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_inspection.py:23: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_inspection.py:23: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_inspection.py:23: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_inspection.py:29: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_lazy.py:37: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_lazy.py:37: Backend.MPARRAY: precision loss SKIPPED [4] tests/main/test_lazy.py:37: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_lazy.py:37: Backend.CUPY: device->host copy SKIPPED [12] tests/main/test_lazy.py:37: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_lazy.py:37: Backend.TORCH_GPU: device->host copy SKIPPED [8] tests/main/test_lazy.py:37: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_lazy.py:37: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_lazy.py:37: Backend.SPARSE: densification SKIPPED [16] tests/main/test_lazy.py:37: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:63: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:63: Backend.MPARRAY: precision loss SKIPPED [1] tests/main/test_lazy.py:63: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_lazy.py:63: Backend.CUPY: device->host copy SKIPPED [3] tests/main/test_lazy.py:63: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:63: Backend.TORCH_GPU: device->host copy SKIPPED [2] tests/main/test_lazy.py:63: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:63: could not import 'sparse': No module named 'sparse' SKIPPED [1] tests/main/test_lazy.py:63: Backend.SPARSE: densification SKIPPED [4] tests/main/test_lazy.py:63: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:76: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:76: Backend.MPARRAY: precision loss SKIPPED [1] tests/main/test_lazy.py:76: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_lazy.py:76: Backend.CUPY: device->host copy SKIPPED [3] tests/main/test_lazy.py:76: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:76: Backend.TORCH_GPU: device->host copy SKIPPED [2] tests/main/test_lazy.py:76: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:76: could not import 'sparse': No module named 'sparse' SKIPPED [1] tests/main/test_lazy.py:76: Backend.SPARSE: densification SKIPPED [4] tests/main/test_lazy.py:76: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:99: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:99: Backend.MPARRAY: precision loss SKIPPED [1] tests/main/test_lazy.py:99: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_lazy.py:99: Backend.CUPY: device->host copy SKIPPED [3] tests/main/test_lazy.py:99: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:99: Backend.TORCH_GPU: device->host copy SKIPPED [2] tests/main/test_lazy.py:99: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:99: could not import 'sparse': No module named 'sparse' SKIPPED [1] tests/main/test_lazy.py:99: Backend.SPARSE: densification SKIPPED [4] tests/main/test_lazy.py:99: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:138: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:162: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:177: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:198: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:219: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:219: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_lazy.py:219: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:219: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:219: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_lazy.py:219: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:266: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [1] tests/main/test_lazy.py:266: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:266: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_lazy.py:266: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:266: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:266: Backend.SPARSE: index by sparse array SKIPPED [1] tests/main/test_lazy.py:266: Backend.JAX: boolean indexing SKIPPED [1] tests/main/test_lazy.py:266: Backend.JAX_GPU: boolean indexing SKIPPED [1] tests/main/test_lazy.py:277: Backend.ARRAY_API_STRICT: device->host copy SKIPPED [1] tests/main/test_lazy.py:277: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:277: Backend.MPARRAY: precision loss SKIPPED [1] tests/main/test_lazy.py:277: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_lazy.py:277: Backend.CUPY: device->host copy SKIPPED [2] tests/main/test_lazy.py:277: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:277: Backend.TORCH: materialize 'meta' device SKIPPED [1] tests/main/test_lazy.py:277: Backend.TORCH_GPU: device->host copy SKIPPED [2] tests/main/test_lazy.py:277: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:277: could not import 'sparse': No module named 'sparse' SKIPPED [1] tests/main/test_lazy.py:277: Backend.SPARSE: densification SKIPPED [4] tests/main/test_lazy.py:277: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:313: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:313: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_lazy.py:313: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:313: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:313: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_lazy.py:313: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:336: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:336: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_lazy.py:336: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:336: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:336: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_lazy.py:336: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:401: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:401: Backend.MPARRAY: precision loss SKIPPED [1] tests/main/test_lazy.py:401: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_lazy.py:401: Backend.CUPY: device->host copy SKIPPED [3] tests/main/test_lazy.py:401: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:401: Backend.TORCH_GPU: device->host copy SKIPPED [2] tests/main/test_lazy.py:401: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:401: could not import 'sparse': No module named 'sparse' SKIPPED [1] tests/main/test_lazy.py:401: Backend.SPARSE: densification SKIPPED [4] tests/main/test_lazy.py:401: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_lazy.py:429: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_lazy.py:429: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_lazy.py:429: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_lazy.py:429: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_lazy.py:429: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_lazy.py:429: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_linalg.py:12: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_linalg.py:12: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_linalg.py:12: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_linalg.py:12: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_linalg.py:12: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_linalg.py:12: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_linalg.py:33: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_linalg.py:33: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_linalg.py:33: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_linalg.py:33: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_linalg.py:33: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_linalg.py:33: could not import 'jax.numpy': No module named 'jax' SKIPPED [6] tests/main/test_linalg.py:40: could not import 'mparray': No module named 'mparray' SKIPPED [6] tests/main/test_linalg.py:40: could not import 'cupy': No module named 'cupy' SKIPPED [12] tests/main/test_linalg.py:40: could not import 'torch': No module named 'torch' SKIPPED [6] tests/main/test_linalg.py:40: could not import 'dask.array': No module named 'dask' SKIPPED [6] tests/main/test_linalg.py:40: could not import 'sparse': No module named 'sparse' SKIPPED [12] tests/main/test_linalg.py:40: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_linalg.py:69: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_linalg.py:69: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_linalg.py:69: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_linalg.py:69: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_linalg.py:69: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_linalg.py:69: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_linalg.py:81: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_linalg.py:81: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_linalg.py:81: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_linalg.py:81: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_linalg.py:81: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_linalg.py:81: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_linalg.py:86: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_linalg.py:86: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_linalg.py:86: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_linalg.py:86: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_linalg.py:86: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_linalg.py:86: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:19: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:19: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:19: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:19: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:19: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:19: could not import 'jax.numpy': No module named 'jax' SKIPPED [5] tests/main/test_manipulation.py:31: could not import 'mparray': No module named 'mparray' SKIPPED [5] tests/main/test_manipulation.py:31: could not import 'cupy': No module named 'cupy' SKIPPED [10] tests/main/test_manipulation.py:31: could not import 'torch': No module named 'torch' SKIPPED [5] tests/main/test_manipulation.py:31: could not import 'dask.array': No module named 'dask' SKIPPED [5] tests/main/test_manipulation.py:31: could not import 'sparse': No module named 'sparse' SKIPPED [10] tests/main/test_manipulation.py:31: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:55: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:55: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:55: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:55: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:55: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:55: could not import 'jax.numpy': No module named 'jax' SKIPPED [5] tests/main/test_manipulation.py:70: could not import 'mparray': No module named 'mparray' SKIPPED [5] tests/main/test_manipulation.py:70: could not import 'cupy': No module named 'cupy' SKIPPED [10] tests/main/test_manipulation.py:70: could not import 'torch': No module named 'torch' SKIPPED [5] tests/main/test_manipulation.py:70: could not import 'dask.array': No module named 'dask' SKIPPED [5] tests/main/test_manipulation.py:70: could not import 'sparse': No module named 'sparse' SKIPPED [10] tests/main/test_manipulation.py:70: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:94: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:94: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:94: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:94: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:94: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:94: could not import 'jax.numpy': No module named 'jax' SKIPPED [5] tests/main/test_manipulation.py:109: could not import 'mparray': No module named 'mparray' SKIPPED [5] tests/main/test_manipulation.py:109: could not import 'cupy': No module named 'cupy' SKIPPED [10] tests/main/test_manipulation.py:109: could not import 'torch': No module named 'torch' SKIPPED [5] tests/main/test_manipulation.py:109: could not import 'dask.array': No module named 'dask' SKIPPED [5] tests/main/test_manipulation.py:109: could not import 'sparse': No module named 'sparse' SKIPPED [10] tests/main/test_manipulation.py:109: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:133: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:133: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:133: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:133: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:133: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:133: could not import 'jax.numpy': No module named 'jax' SKIPPED [5] tests/main/test_manipulation.py:148: could not import 'mparray': No module named 'mparray' SKIPPED [5] tests/main/test_manipulation.py:148: could not import 'cupy': No module named 'cupy' SKIPPED [10] tests/main/test_manipulation.py:148: could not import 'torch': No module named 'torch' SKIPPED [5] tests/main/test_manipulation.py:148: could not import 'dask.array': No module named 'dask' SKIPPED [5] tests/main/test_manipulation.py:148: could not import 'sparse': No module named 'sparse' SKIPPED [10] tests/main/test_manipulation.py:148: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:172: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:172: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:172: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:172: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:172: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:172: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:214: Backend.NUMPY_READONLY: xp=xp SKIPPED [1] tests/main/test_manipulation.py:214: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:214: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:214: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:214: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:214: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:214: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:276: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:276: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:276: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:276: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:276: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:276: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:283: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:283: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:283: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:283: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:283: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:283: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:290: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:290: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:290: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:290: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:290: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:290: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:303: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:303: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:303: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:303: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:303: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:303: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:308: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:308: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:308: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:308: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:308: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:308: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:314: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:314: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:314: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:314: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:314: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:314: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:318: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:318: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:318: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:318: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:318: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:318: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:325: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:325: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:325: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:325: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:325: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:325: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:330: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:330: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:330: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:330: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:330: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:330: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:338: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:338: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:338: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:338: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:338: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:338: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:343: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:343: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:343: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:343: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:343: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:343: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:348: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:348: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:348: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:348: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:348: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:348: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:352: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:352: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:352: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:352: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:352: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:352: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:356: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:356: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:356: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:356: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:356: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:356: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_manipulation.py:367: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_manipulation.py:367: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_manipulation.py:367: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_manipulation.py:367: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_manipulation.py:367: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_manipulation.py:367: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_searching.py:149: Backend.MPARRAY: see mdhaber/mparray#20 SKIPPED [1] tests/main/test_searching.py:149: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_searching.py:149: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_searching.py:149: Backend.DASK: no take_along_axis SKIPPED [1] tests/main/test_searching.py:149: Backend.SPARSE: no searchsorted SKIPPED [2] tests/main/test_searching.py:149: could not import 'jax.numpy': No module named 'jax' SKIPPED [128] tests/main/test_searching.py:181: Redundant; `xpx_searchsorted` delegates to `_funcs_searchsorted` for multidimensional input. SKIPPED [352] tests/main/test_searching.py:154: Backend.MPARRAY: see mdhaber/mparray#20 SKIPPED [352] tests/main/test_searching.py:154: could not import 'cupy': No module named 'cupy' SKIPPED [704] tests/main/test_searching.py:154: could not import 'torch': No module named 'torch' SKIPPED [352] tests/main/test_searching.py:154: Backend.DASK: no take_along_axis SKIPPED [352] tests/main/test_searching.py:154: Backend.SPARSE: no searchsorted SKIPPED [704] tests/main/test_searching.py:154: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:17: Backend.ARRAY_API_STRICT: array-agnostic fallback SKIPPED [1] tests/main/test_set.py:17: Backend.ARRAY_API_STRICTEST: array-agnostic fallback SKIPPED [1] tests/main/test_set.py:17: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:17: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:17: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:17: Backend.DASK: array-agnostic fallback SKIPPED [1] tests/main/test_set.py:17: Backend.SPARSE: array-agnostic fallback SKIPPED [2] tests/main/test_set.py:17: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:38: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:38: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:38: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:38: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:38: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:38: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:42: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:42: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:42: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:42: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:42: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:42: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:46: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:46: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:46: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:46: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:46: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:46: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:50: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:50: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:50: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:50: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:50: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:50: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:54: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:54: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:54: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:54: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:54: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:54: could not import 'jax.numpy': No module named 'jax' SKIPPED [3] tests/main/test_set.py:63: could not import 'mparray': No module named 'mparray' SKIPPED [3] tests/main/test_set.py:63: could not import 'cupy': No module named 'cupy' SKIPPED [6] tests/main/test_set.py:63: could not import 'torch': No module named 'torch' SKIPPED [3] tests/main/test_set.py:63: could not import 'dask.array': No module named 'dask' SKIPPED [3] tests/main/test_set.py:63: could not import 'sparse': No module named 'sparse' SKIPPED [6] tests/main/test_set.py:63: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:68: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:68: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:68: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:68: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:68: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:68: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:90: Backend.ARRAY_API_STRICTEST: no unique_values SKIPPED [1] tests/main/test_set.py:90: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:90: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:90: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:90: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:90: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:90: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:118: Backend.ARRAY_API_STRICTEST: no unique_values SKIPPED [1] tests/main/test_set.py:118: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:118: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:118: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:118: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:118: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:118: could not import 'jax.numpy': No module named 'jax' SKIPPED [18] tests/main/test_set.py:125: Backend.ARRAY_API_STRICTEST: no unique_values SKIPPED [18] tests/main/test_set.py:125: could not import 'mparray': No module named 'mparray' SKIPPED [18] tests/main/test_set.py:125: could not import 'cupy': No module named 'cupy' SKIPPED [36] tests/main/test_set.py:125: could not import 'torch': No module named 'torch' SKIPPED [18] tests/main/test_set.py:125: could not import 'dask.array': No module named 'dask' SKIPPED [18] tests/main/test_set.py:125: could not import 'sparse': No module named 'sparse' SKIPPED [36] tests/main/test_set.py:125: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_set.py:141: Backend.ARRAY_API_STRICTEST: no unique_values SKIPPED [2] tests/main/test_set.py:141: Backend.NUMPY_READONLY: xp=xp SKIPPED [2] tests/main/test_set.py:141: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_set.py:141: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_set.py:141: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_set.py:141: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_set.py:141: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_set.py:141: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_set.py:163: Backend.ARRAY_API_STRICTEST: no unique_values SKIPPED [2] tests/main/test_set.py:163: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_set.py:163: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:163: Backend.TORCH: device='meta' does not support unknown shapes SKIPPED [2] tests/main/test_set.py:163: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_set.py:163: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_set.py:163: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_set.py:163: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:172: Backend.ARRAY_API_STRICTEST: no unique_values SKIPPED [1] tests/main/test_set.py:172: Backend.NUMPY_READONLY: xp=xp SKIPPED [1] tests/main/test_set.py:172: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:172: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:172: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:172: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:172: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:172: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:183: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:183: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:183: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:183: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:183: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:183: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:201: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:201: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:201: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:201: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:201: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:201: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:209: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:209: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:209: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:209: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:209: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:209: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:222: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:222: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:222: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:222: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:222: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:222: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:238: Backend.ARRAY_API_STRICTEST: data_dependent_shapes flag for unique_values is disabled SKIPPED [1] tests/main/test_set.py:238: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:238: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:238: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:238: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:238: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:238: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:245: Backend.ARRAY_API_STRICTEST: data_dependent_shapes flag for unique_values is disabled SKIPPED [1] tests/main/test_set.py:245: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:245: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:245: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:245: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:245: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:245: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:252: Backend.ARRAY_API_STRICTEST: data_dependent_shapes flag for unique_values is disabled SKIPPED [1] tests/main/test_set.py:252: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:252: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_set.py:252: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:252: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:252: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:252: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_set.py:259: Backend.ARRAY_API_STRICTEST: data_dependent_shapes flag for unique_values is disabled SKIPPED [1] tests/main/test_set.py:259: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_set.py:259: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_set.py:259: Backend.TORCH: materialize 'meta' device SKIPPED [1] tests/main/test_set.py:259: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_set.py:259: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_set.py:259: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_set.py:259: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_sorting.py:88: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_sorting.py:88: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_sorting.py:88: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_sorting.py:88: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_sorting.py:88: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_sorting.py:88: could not import 'jax.numpy': No module named 'jax' SKIPPED [3] tests/main/test_sorting.py:91: could not import 'mparray': No module named 'mparray' SKIPPED [3] tests/main/test_sorting.py:91: could not import 'cupy': No module named 'cupy' SKIPPED [6] tests/main/test_sorting.py:91: could not import 'torch': No module named 'torch' SKIPPED [3] tests/main/test_sorting.py:91: could not import 'dask.array': No module named 'dask' SKIPPED [3] tests/main/test_sorting.py:91: could not import 'sparse': No module named 'sparse' SKIPPED [6] tests/main/test_sorting.py:91: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_sorting.py:95: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_sorting.py:95: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_sorting.py:95: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_sorting.py:95: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_sorting.py:95: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_sorting.py:95: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_sorting.py:142: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_sorting.py:142: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_sorting.py:142: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_sorting.py:142: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_sorting.py:142: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_sorting.py:142: could not import 'jax.numpy': No module named 'jax' SKIPPED [3] tests/main/test_sorting.py:146: could not import 'mparray': No module named 'mparray' SKIPPED [3] tests/main/test_sorting.py:146: could not import 'cupy': No module named 'cupy' SKIPPED [6] tests/main/test_sorting.py:146: could not import 'torch': No module named 'torch' SKIPPED [3] tests/main/test_sorting.py:146: could not import 'dask.array': No module named 'dask' SKIPPED [3] tests/main/test_sorting.py:146: could not import 'sparse': No module named 'sparse' SKIPPED [6] tests/main/test_sorting.py:146: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_sorting.py:151: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_sorting.py:151: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_sorting.py:151: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_sorting.py:151: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_sorting.py:151: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_sorting.py:151: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:21: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:21: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:21: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:21: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:21: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:21: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:27: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:27: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:27: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:27: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:27: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:27: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:32: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:32: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:32: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:32: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:32: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:32: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:53: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:53: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:53: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:53: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:53: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:53: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:70: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:70: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:70: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:70: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:70: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:70: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:79: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:79: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:79: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:79: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:79: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:79: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:86: Backend.NUMPY_READONLY: xp=xp SKIPPED [1] tests/main/test_statistical.py:86: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:86: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:86: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:86: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:86: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:86: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:96: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:96: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:96: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:96: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:96: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:96: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_statistical.py:106: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_statistical.py:106: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_statistical.py:106: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_statistical.py:106: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_statistical.py:106: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_statistical.py:106: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_statistical.py:119: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_statistical.py:119: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_statistical.py:119: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_statistical.py:119: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_statistical.py:119: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_statistical.py:119: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:130: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:130: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:130: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:130: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:130: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:130: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:138: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:138: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:138: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:138: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:138: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:138: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:149: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:149: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:149: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:149: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:149: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:149: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:158: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:158: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:158: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:158: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:158: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:158: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:166: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:166: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:166: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:166: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:166: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:166: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:174: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:174: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:174: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:174: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:174: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:174: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:189: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:189: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:189: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:189: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:189: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:189: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:200: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:200: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:200: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:200: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:200: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:200: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:216: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:216: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:216: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:216: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:216: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:216: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:221: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:221: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:221: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:221: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:221: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:221: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:231: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:231: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:231: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:231: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:231: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:231: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:239: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_statistical.py:268: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:288: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:288: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:288: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:288: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:288: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:288: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:304: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:304: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:304: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:304: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:304: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:304: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:321: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:321: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:321: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:321: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:321: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:321: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:332: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:332: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:332: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:332: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:332: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:332: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:336: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:336: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:336: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:336: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:336: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:336: could not import 'jax.numpy': No module named 'jax' SKIPPED [3] tests/main/test_statistical.py:349: could not import 'mparray': No module named 'mparray' SKIPPED [3] tests/main/test_statistical.py:349: could not import 'cupy': No module named 'cupy' SKIPPED [3] tests/main/test_statistical.py:349: Backend.TORCH: torch.nanmin does not support tensors on meta device SKIPPED [3] tests/main/test_statistical.py:349: could not import 'torch': No module named 'torch' SKIPPED [3] tests/main/test_statistical.py:349: could not import 'dask.array': No module named 'dask' SKIPPED [3] tests/main/test_statistical.py:349: could not import 'sparse': No module named 'sparse' SKIPPED [6] tests/main/test_statistical.py:349: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_statistical.py:358: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_statistical.py:358: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_statistical.py:358: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_statistical.py:358: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_statistical.py:358: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_statistical.py:358: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:369: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:369: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:369: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:369: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:369: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:369: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:385: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:385: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:385: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:385: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:385: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:385: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:402: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:402: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:402: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:402: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:402: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:402: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:413: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:413: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:413: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:413: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:413: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:413: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:417: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:417: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:417: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:417: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:417: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:417: could not import 'jax.numpy': No module named 'jax' SKIPPED [3] tests/main/test_statistical.py:430: could not import 'mparray': No module named 'mparray' SKIPPED [3] tests/main/test_statistical.py:430: could not import 'cupy': No module named 'cupy' SKIPPED [3] tests/main/test_statistical.py:430: Backend.TORCH: torch.nanmax does not support tensors on meta device SKIPPED [3] tests/main/test_statistical.py:430: could not import 'torch': No module named 'torch' SKIPPED [3] tests/main/test_statistical.py:430: could not import 'dask.array': No module named 'dask' SKIPPED [3] tests/main/test_statistical.py:430: could not import 'sparse': No module named 'sparse' SKIPPED [6] tests/main/test_statistical.py:430: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_statistical.py:439: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_statistical.py:439: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_statistical.py:439: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_statistical.py:439: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_statistical.py:439: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_statistical.py:439: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:450: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:450: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:450: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:450: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:450: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:450: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:465: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:465: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:465: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:465: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:465: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:465: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:482: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:482: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:482: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:482: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:482: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:482: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:489: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:489: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:489: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:489: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:489: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:489: could not import 'jax.numpy': No module named 'jax' SKIPPED [3] tests/main/test_statistical.py:493: could not import 'mparray': No module named 'mparray' SKIPPED [3] tests/main/test_statistical.py:493: could not import 'cupy': No module named 'cupy' SKIPPED [3] tests/main/test_statistical.py:493: Backend.TORCH: torch.nansum does not support tensors on meta device SKIPPED [3] tests/main/test_statistical.py:493: could not import 'torch': No module named 'torch' SKIPPED [3] tests/main/test_statistical.py:493: could not import 'dask.array': No module named 'dask' SKIPPED [3] tests/main/test_statistical.py:493: could not import 'sparse': No module named 'sparse' SKIPPED [6] tests/main/test_statistical.py:493: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_statistical.py:502: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_statistical.py:502: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_statistical.py:502: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_statistical.py:502: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_statistical.py:502: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_statistical.py:502: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:513: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:513: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:513: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:513: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:513: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:513: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:527: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:527: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:527: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:527: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:527: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:527: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:544: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:544: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:544: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:544: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:544: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:544: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_statistical.py:552: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_statistical.py:552: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_statistical.py:552: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_statistical.py:552: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_statistical.py:552: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_statistical.py:552: could not import 'jax.numpy': No module named 'jax' SKIPPED [3] tests/main/test_statistical.py:556: could not import 'mparray': No module named 'mparray' SKIPPED [3] tests/main/test_statistical.py:556: could not import 'cupy': No module named 'cupy' SKIPPED [3] tests/main/test_statistical.py:556: Backend.TORCH: torch.nanmean does not support tensors on meta device SKIPPED [3] tests/main/test_statistical.py:556: could not import 'torch': No module named 'torch' SKIPPED [3] tests/main/test_statistical.py:556: could not import 'dask.array': No module named 'dask' SKIPPED [3] tests/main/test_statistical.py:556: could not import 'sparse': No module named 'sparse' SKIPPED [6] tests/main/test_statistical.py:556: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_statistical.py:565: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_statistical.py:565: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_statistical.py:565: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_statistical.py:565: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_statistical.py:565: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_statistical.py:565: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:31: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:31: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:31: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:31: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:31: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:31: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:36: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:36: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_testing.py:36: Backend.TORCH: materialize 'meta' device SKIPPED [1] tests/main/test_testing.py:36: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:36: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:36: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:36: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_testing.py:47: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_testing.py:47: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_testing.py:47: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_testing.py:47: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_testing.py:47: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_testing.py:47: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_testing.py:60: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_testing.py:60: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_testing.py:60: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_testing.py:60: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_testing.py:60: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_testing.py:60: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_testing.py:70: Backend.NUMPY: test other ns vs. numpy SKIPPED [4] tests/main/test_testing.py:70: Backend.NUMPY_READONLY: test other ns vs. numpy SKIPPED [4] tests/main/test_testing.py:70: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_testing.py:70: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_testing.py:70: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_testing.py:70: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_testing.py:70: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_testing.py:70: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_testing.py:95: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_testing.py:95: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_testing.py:95: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_testing.py:95: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_testing.py:95: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_testing.py:95: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_testing.py:112: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_testing.py:112: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_testing.py:112: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_testing.py:112: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_testing.py:112: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_testing.py:112: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_testing.py:126: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_testing.py:126: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_testing.py:126: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_testing.py:126: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_testing.py:126: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_testing.py:126: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_testing.py:148: could not import 'mparray': No module named 'mparray' SKIPPED [2] tests/main/test_testing.py:148: could not import 'cupy': No module named 'cupy' SKIPPED [4] tests/main/test_testing.py:148: could not import 'torch': No module named 'torch' SKIPPED [2] tests/main/test_testing.py:148: could not import 'dask.array': No module named 'dask' SKIPPED [2] tests/main/test_testing.py:148: could not import 'sparse': No module named 'sparse' SKIPPED [4] tests/main/test_testing.py:148: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:164: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:164: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:164: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:164: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:164: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:164: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_testing.py:170: Backend.ARRAY_API_STRICTEST: boolean indexing SKIPPED [4] tests/main/test_testing.py:170: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_testing.py:170: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_testing.py:170: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_testing.py:170: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_testing.py:170: Backend.SPARSE: index by sparse array SKIPPED [8] tests/main/test_testing.py:170: could not import 'jax.numpy': No module named 'jax' SKIPPED [4] tests/main/test_testing.py:203: could not import 'mparray': No module named 'mparray' SKIPPED [4] tests/main/test_testing.py:203: could not import 'cupy': No module named 'cupy' SKIPPED [8] tests/main/test_testing.py:203: could not import 'torch': No module named 'torch' SKIPPED [4] tests/main/test_testing.py:203: could not import 'dask.array': No module named 'dask' SKIPPED [4] tests/main/test_testing.py:203: could not import 'sparse': No module named 'sparse' SKIPPED [8] tests/main/test_testing.py:203: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:221: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:221: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:221: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:221: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:221: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:221: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:292: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:292: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:292: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:292: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:292: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:292: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:336: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:336: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:336: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:336: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:336: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:336: could not import 'jax.numpy': No module named 'jax' SKIPPED [2] tests/main/test_testing.py:368: could not import 'scipy': No module named 'scipy' SKIPPED [1] tests/main/test_testing.py:364: Backend.MPARRAY: mparray lacks __array_ufunc__ SKIPPED [1] tests/main/test_testing.py:364: could not import 'cupy': No module named 'cupy' SKIPPED [1] tests/main/test_testing.py:364: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:364: Backend.TORCH_GPU: device->host copy SKIPPED [1] tests/main/test_testing.py:364: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:364: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:364: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:451: Backend.MPARRAY: precision loss SKIPPED [1] tests/main/test_testing.py:451: Backend.CUPY: converts to NumPy SKIPPED [1] tests/main/test_testing.py:451: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:451: Backend.TORCH_GPU: converts to NumPy SKIPPED [1] tests/main/test_testing.py:451: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:451: Backend.SPARSE: converts to NumPy SKIPPED [1] tests/main/test_testing.py:451: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:451: Backend.JAX_GPU: converts to NumPy SKIPPED [1] tests/main/test_testing.py:470: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:470: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:470: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:470: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:470: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:470: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:478: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:478: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:478: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:478: Backend.DASK: calls dask.compute() SKIPPED [1] tests/main/test_testing.py:478: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:478: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:491: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:491: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:491: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:491: Backend.DASK: calls dask.compute() SKIPPED [1] tests/main/test_testing.py:491: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:491: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:502: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:502: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:502: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:502: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:502: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:502: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:522: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:555: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:555: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:555: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:555: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:555: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:555: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:566: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:587: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:587: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:587: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:587: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:587: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:587: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:619: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:619: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:619: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:619: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:619: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:619: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/main/test_testing.py:640: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:665: could not import 'mparray': No module named 'mparray' SKIPPED [1] tests/main/test_testing.py:665: could not import 'cupy': No module named 'cupy' SKIPPED [2] tests/main/test_testing.py:665: could not import 'torch': No module named 'torch' SKIPPED [1] tests/main/test_testing.py:665: could not import 'dask.array': No module named 'dask' SKIPPED [1] tests/main/test_testing.py:665: could not import 'sparse': No module named 'sparse' SKIPPED [2] tests/main/test_testing.py:665: could not import 'jax.numpy': No module named 'jax' SKIPPED [1] tests/run_deps/test_run_deps.py:13: Test for when `numpy` is not importable. ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-set-40.0-expect_list0-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-set-40.0-expect_list0-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-set-40.0-expect_list0-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-add-40.0-expect_list1-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-add-40.0-expect_list1-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-add-40.0-expect_list1-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-subtract-100.0-expect_list2-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-subtract-100.0-expect_list2-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-subtract-100.0-expect_list2-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-multiply-2.0-expect_list3-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-multiply-2.0-expect_list3-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-multiply-2.0-expect_list3-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-divide-2.0-expect_list4-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-divide-2.0-expect_list4-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-divide-2.0-expect_list4-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-power-2.0-expect_list5-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-power-2.0-expect_list5-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-power-2.0-expect_list5-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-min-25.0-expect_list6-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-min-25.0-expect_list6-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-min-25.0-expect_list6-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-max-25.0-expect_list7-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-max-25.0-expect_list7-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-0-max-25.0-expect_list7-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-set-40.0-expect_list0-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-set-40.0-expect_list0-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-set-40.0-expect_list0-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-add-40.0-expect_list1-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-add-40.0-expect_list1-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-add-40.0-expect_list1-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-subtract-100.0-expect_list2-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-subtract-100.0-expect_list2-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-subtract-100.0-expect_list2-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-multiply-2.0-expect_list3-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-multiply-2.0-expect_list3-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-multiply-2.0-expect_list3-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-divide-2.0-expect_list4-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-divide-2.0-expect_list4-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-divide-2.0-expect_list4-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-power-2.0-expect_list5-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-power-2.0-expect_list5-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-power-2.0-expect_list5-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-min-25.0-expect_list6-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-min-25.0-expect_list6-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-min-25.0-expect_list6-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-max-25.0-expect_list7-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-max-25.0-expect_list7-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-1-1-max-25.0-expect_list7-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-set-40.0-expect_list0-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-set-40.0-expect_list0-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-set-40.0-expect_list0-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-add-40.0-expect_list1-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-add-40.0-expect_list1-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-add-40.0-expect_list1-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-subtract-100.0-expect_list2-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-subtract-100.0-expect_list2-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-subtract-100.0-expect_list2-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-multiply-2.0-expect_list3-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-multiply-2.0-expect_list3-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-multiply-2.0-expect_list3-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-divide-2.0-expect_list4-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-divide-2.0-expect_list4-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-divide-2.0-expect_list4-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-power-2.0-expect_list5-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-power-2.0-expect_list5-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-power-2.0-expect_list5-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-min-25.0-expect_list6-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-min-25.0-expect_list6-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-min-25.0-expect_list6-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-max-25.0-expect_list7-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-max-25.0-expect_list7-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-0-max-25.0-expect_list7-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-set-40.0-expect_list0-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-set-40.0-expect_list0-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-set-40.0-expect_list0-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-add-40.0-expect_list1-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-add-40.0-expect_list1-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-add-40.0-expect_list1-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-subtract-100.0-expect_list2-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-subtract-100.0-expect_list2-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-subtract-100.0-expect_list2-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-multiply-2.0-expect_list3-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-multiply-2.0-expect_list3-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-multiply-2.0-expect_list3-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-divide-2.0-expect_list4-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-divide-2.0-expect_list4-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-divide-2.0-expect_list4-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-power-2.0-expect_list5-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-power-2.0-expect_list5-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-power-2.0-expect_list5-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-min-25.0-expect_list6-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-min-25.0-expect_list6-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-min-25.0-expect_list6-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-max-25.0-expect_list7-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-max-25.0-expect_list7-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-1-1-max-25.0-expect_list7-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-set-40.0-expect_list0-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-set-40.0-expect_list0-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-set-40.0-expect_list0-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-add-40.0-expect_list1-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-add-40.0-expect_list1-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-add-40.0-expect_list1-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-subtract-100.0-expect_list2-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-subtract-100.0-expect_list2-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-subtract-100.0-expect_list2-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-multiply-2.0-expect_list3-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-multiply-2.0-expect_list3-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-multiply-2.0-expect_list3-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-divide-2.0-expect_list4-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-divide-2.0-expect_list4-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-divide-2.0-expect_list4-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-power-2.0-expect_list5-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-power-2.0-expect_list5-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-power-2.0-expect_list5-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-min-25.0-expect_list6-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-min-25.0-expect_list6-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-min-25.0-expect_list6-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-max-25.0-expect_list7-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-max-25.0-expect_list7-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-False-0-0-max-25.0-expect_list7-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-set-40.0-expect_list0-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-set-40.0-expect_list0-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-set-40.0-expect_list0-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-add-40.0-expect_list1-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-add-40.0-expect_list1-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-add-40.0-expect_list1-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-subtract-100.0-expect_list2-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-subtract-100.0-expect_list2-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-subtract-100.0-expect_list2-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-multiply-2.0-expect_list3-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-multiply-2.0-expect_list3-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-multiply-2.0-expect_list3-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-divide-2.0-expect_list4-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-divide-2.0-expect_list4-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-divide-2.0-expect_list4-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-power-2.0-expect_list5-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-power-2.0-expect_list5-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-power-2.0-expect_list5-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-min-25.0-expect_list6-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-min-25.0-expect_list6-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-min-25.0-expect_list6-None] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-max-25.0-expect_list7-False] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-max-25.0-expect_list7-True] ERROR tests/main/test_at.py::test_update_ops[array_api_strict-True-0-0-max-25.0-expect_list7-None] ERROR tests/main/test_at.py::test_copy_default[array_api_strict-set] - ValueE... ERROR tests/main/test_at.py::test_copy_default[array_api_strict-add] - ValueE... ERROR tests/main/test_at.py::test_copy_default[array_api_strict-subtract] - V... ERROR tests/main/test_at.py::test_copy_default[array_api_strict-multiply] - V... ERROR tests/main/test_at.py::test_copy_default[array_api_strict-divide] - Val... ERROR tests/main/test_at.py::test_copy_default[array_api_strict-power] - Valu... ERROR tests/main/test_at.py::test_copy_default[array_api_strict-min] - ValueE... ERROR tests/main/test_at.py::test_copy_default[array_api_strict-max] - ValueE... ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-set-False-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-set-False-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-set-True-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-set-True-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-add-False-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-add-False-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-add-True-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-add-True-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-subtract-False-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-subtract-False-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-subtract-True-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-subtract-True-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-multiply-False-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-multiply-False-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-multiply-True-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-multiply-True-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-divide-False-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-divide-False-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-divide-True-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-divide-True-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-power-False-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-power-False-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-power-True-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-power-True-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-min-False-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-min-False-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-min-True-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-min-True-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-max-False-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-max-False-None] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-max-True-True] ERROR tests/main/test_at.py::test_incompatible_dtype[array_api_strict-max-True-None] ERROR tests/main/test_at.py::test_bool_mask_nd[array_api_strict] - ValueError... ERROR tests/main/test_at.py::test_no_inf_warnings[array_api_strict-False] - V... ERROR tests/main/test_at.py::test_no_inf_warnings[array_api_strict-True] - Va... ERROR tests/main/test_at.py::test_gh134[array_api_strict-False-None] - ValueE... ERROR tests/main/test_at.py::test_gh134[array_api_strict-False-False] - Value... ERROR tests/main/test_at.py::test_gh134[array_api_strict-True-None] - ValueEr... ERROR tests/main/test_at.py::test_gh134[array_api_strict-True-False] - ValueE... ERROR tests/main/test_at.py::test_device[array_api_strict] - ValueError: Unsu... ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-1-0] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-1-1] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-1-2] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-1-3] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-3-0] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-3-1] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-3-2] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-3-3] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-10-0] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-10-1] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-10-2] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strict-10-3] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-1-0] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-1-1] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-1-2] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-1-3] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-3-0] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-3-1] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-3-2] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-3-3] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-10-0] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-10-1] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-10-2] ERROR tests/main/test_creation.py::TestOneHot::test_dims_and_classes[array_api_strictest-10-3] ERROR tests/main/test_creation.py::TestOneHot::test_basic[array_api_strict] ERROR tests/main/test_creation.py::TestOneHot::test_basic[array_api_strictest] ERROR tests/main/test_creation.py::TestOneHot::test_2d[array_api_strict] - Va... ERROR tests/main/test_creation.py::TestOneHot::test_2d[array_api_strictest] ERROR tests/main/test_creation.py::TestOneHot::test_abstract_size[array_api_strict] ERROR tests/main/test_creation.py::TestOneHot::test_out_of_bound[array_api_strict] ERROR tests/main/test_creation.py::TestOneHot::test_out_of_bound[array_api_strictest] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strict-int8] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strict-int16] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strict-int32] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strict-int64] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strict-uint8] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strict-uint16] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strict-uint32] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strict-uint64] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strictest-int8] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strictest-int16] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strictest-int32] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strictest-int64] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strictest-uint8] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strictest-uint16] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strictest-uint32] ERROR tests/main/test_creation.py::TestOneHot::test_int_types[array_api_strictest-uint64] ERROR tests/main/test_creation.py::TestOneHot::test_custom_dtype[array_api_strict] ERROR tests/main/test_creation.py::TestOneHot::test_custom_dtype[array_api_strictest] ERROR tests/main/test_creation.py::TestOneHot::test_axis[array_api_strict] - ... ERROR tests/main/test_creation.py::TestOneHot::test_axis[array_api_strictest] ERROR tests/main/test_creation.py::TestOneHot::test_non_integer[array_api_strict] ERROR tests/main/test_creation.py::TestOneHot::test_non_integer[array_api_strictest] ERROR tests/main/test_creation.py::TestOneHot::test_device[array_api_strict] ERROR tests/main/test_creation.py::TestOneHot::test_device[array_api_strictest] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_numpy[array_api_strict] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_numpy[array_api_strictest] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-1-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-2-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-3-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-4-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-5-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-6-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-7-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-8-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strict-9-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-1-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-2-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-3-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-4-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-5-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-6-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-7-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-8-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_1d_from_scipy[array_api_strictest-9-9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_0d_raises[array_api_strict] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_0d_raises[array_api_strictest] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape0] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strict-shape9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape0] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape1] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape2] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape3] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape4] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape5] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape6] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape7] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape8] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_nd[array_api_strictest-shape9] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_device[array_api_strict] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_device[array_api_strictest] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_xp[array_api_strict] ERROR tests/main/test_creation.py::TestCreateDiagonal::test_xp[array_api_strictest] ERROR tests/main/test_deprecation.py::TestDeprecatedFunctions::test_broadcast_shapes[array_api_strict] ERROR tests/main/test_deprecation.py::TestDeprecatedFunctions::test_broadcast_shapes[array_api_strictest] ERROR tests/main/test_deprecation.py::TestDeprecatedFunctions::test_expand_dims[array_api_strict] ERROR tests/main/test_deprecation.py::TestDeprecatedFunctions::test_expand_dims[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_f1_f2[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_f1_f2[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_fill_value[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_fill_value[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_args_tuple[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_args_tuple[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_broadcast[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_broadcast[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation_fill_value[array_api_strict-int32-int32-3] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation_fill_value[array_api_strict-int32-int32-fill_value_raw1] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation_fill_value[array_api_strict-int8-int16-3] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation_fill_value[array_api_strict-int8-int16-fill_value_raw1] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation_fill_value[array_api_strictest-int32-int32-3] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation_fill_value[array_api_strictest-int32-int32-fill_value_raw1] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation_fill_value[array_api_strictest-int8-int16-3] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dtype_propagation_fill_value[array_api_strictest-int8-int16-fill_value_raw1] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dont_overwrite_fill_value[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dont_overwrite_fill_value[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_dont_run_on_false[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_bad_args[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_bad_args[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_xp[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_xp[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_device[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_device[array_api_strictest] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_hypothesis[array_api_strict] ERROR tests/main/test_elementwise.py::TestApplyWhere::test_hypothesis[array_api_strictest] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-0.0-0.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-0.0-0.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1.0-1.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1.0-1.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1.0-2.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1.0-2.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1.0--1.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1.0--1.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-100.0-101.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-100.0-101.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-0-0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-0-0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1-1-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1-1-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1-2-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1-2-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1--1-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1--1-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-(1+1j)-(1+1j)-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-(1+1j)-(1+1j)-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-(1+1j)-(1-1j)-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-(1+1j)-(1-1j)-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-inf-inf-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-inf-inf-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-inf-100.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-inf-100.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-inf--inf-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-inf--inf-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict--inf--inf-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict--inf--inf-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-nan-nan-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-nan-nan-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-nan-100.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-nan-100.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1000000.0-1000001.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1000000.0-1000001.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1000000.0-1000100.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1000000.0-1000100.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1e-06-1.1e-06-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1e-06-1.1e-06-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1e-07-1.1e-07-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-1e-07-1.1e-07-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-(1000000+0j)-(1000000+1j)-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-(1000000+0j)-(1000000+1j)-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-(1000000+0j)-(1000000+100j)-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strict-(1000000+0j)-(1000000+100j)-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-0.0-0.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-0.0-0.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1.0-1.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1.0-1.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1.0-2.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1.0-2.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1.0--1.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1.0--1.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-100.0-101.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-100.0-101.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-0-0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-0-0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1-1-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1-1-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1-2-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1-2-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1--1-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1--1-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-(1+1j)-(1+1j)-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-(1+1j)-(1+1j)-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-(1+1j)-(1-1j)-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-(1+1j)-(1-1j)-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-inf-inf-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-inf-inf-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-inf-100.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-inf-100.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-inf--inf-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-inf--inf-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest--inf--inf-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest--inf--inf-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-nan-nan-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-nan-nan-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-nan-100.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-nan-100.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1000000.0-1000001.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1000000.0-1000001.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1000000.0-1000100.0-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1000000.0-1000100.0-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1e-06-1.1e-06-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1e-06-1.1e-06-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1e-07-1.1e-07-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-1e-07-1.1e-07-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-(1000000+0j)-(1000000+1j)-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-(1000000+0j)-(1000000+1j)-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-(1000000+0j)-(1000000+100j)-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_basic[array_api_strictest-(1000000+0j)-(1000000+100j)-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_broadcast[array_api_strict-float32] ERROR tests/main/test_elementwise.py::TestIsClose::test_broadcast[array_api_strict-int32] ERROR tests/main/test_elementwise.py::TestIsClose::test_broadcast[array_api_strictest-float32] ERROR tests/main/test_elementwise.py::TestIsClose::test_broadcast[array_api_strictest-int32] ERROR tests/main/test_elementwise.py::TestIsClose::test_some_inf[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_some_inf[array_api_strictest] ERROR tests/main/test_elementwise.py::TestIsClose::test_equal_nan[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_equal_nan[array_api_strictest] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance[array_api_strict-float32] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance[array_api_strict-complex64] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance[array_api_strict-int32] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance[array_api_strictest-float32] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance[array_api_strictest-complex64] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance[array_api_strictest-int32] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance_integer_overflow[array_api_strict-int8] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance_integer_overflow[array_api_strict-uint8] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance_integer_overflow[array_api_strictest-int8] ERROR tests/main/test_elementwise.py::TestIsClose::test_tolerance_integer_overflow[array_api_strictest-uint8] ERROR tests/main/test_elementwise.py::TestIsClose::test_very_small_numbers[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_very_small_numbers[array_api_strictest] ERROR tests/main/test_elementwise.py::TestIsClose::test_bool_dtype[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_bool_dtype[array_api_strictest] ERROR tests/main/test_elementwise.py::TestIsClose::test_none_shape[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_none_shape_bool[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_python_scalar[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_python_scalar[array_api_strictest] ERROR tests/main/test_elementwise.py::TestIsClose::test_xp[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_xp[array_api_strictest] ERROR tests/main/test_elementwise.py::TestIsClose::test_device[array_api_strict-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_device[array_api_strict-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_device[array_api_strictest-True] ERROR tests/main/test_elementwise.py::TestIsClose::test_device[array_api_strictest-False] ERROR tests/main/test_elementwise.py::TestIsClose::test_array_on_device_with_scalar[array_api_strict] ERROR tests/main/test_elementwise.py::TestIsClose::test_array_on_device_with_scalar[array_api_strictest] ERROR tests/main/test_elementwise.py::TestNanToNum::test_bool[array_api_strict] ERROR tests/main/test_elementwise.py::TestNanToNum::test_bool[array_api_strictest] ERROR tests/main/test_elementwise.py::TestNanToNum::test_scalar_pos_inf[array_api_strict] ERROR tests/main/test_elementwise.py::TestNanToNum::test_scalar_pos_inf[array_api_strictest] ERROR tests/main/test_elementwise.py::TestNanToNum::test_scalar_neg_inf[array_api_strict] ERROR tests/main/test_elementwise.py::TestNanToNum::test_scalar_neg_inf[array_api_strictest] ERROR tests/main/test_elementwise.py::TestNanToNum::test_scalar_nan[array_api_strict] ERROR tests/main/test_elementwise.py::TestNanToNum::test_scalar_nan[array_api_strictest] ERROR tests/main/test_elementwise.py::TestNanToNum::test_real[array_api_strict] ERROR tests/main/test_elementwise.py::TestNanToNum::test_real[array_api_strictest] ERROR tests/main/test_elementwise.py::TestNanToNum::test_complex[array_api_strict] ERROR tests/main/test_elementwise.py::TestNanToNum::test_complex[array_api_strictest] ERROR tests/main/test_elementwise.py::TestNanToNum::test_empty_array[array_api_strict] ERROR tests/main/test_elementwise.py::TestNanToNum::test_empty_array[array_api_strictest] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strict-in_vals0-3-out_vals0] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strict-in_vals1-3.0-out_vals1] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strict-in_vals2-3-out_vals2] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strict-in_vals3-3.0-out_vals3] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strict-in_vals4-3.0-out_vals4] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strictest-in_vals0-3-out_vals0] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strictest-in_vals1-3.0-out_vals1] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strictest-in_vals2-3-out_vals2] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strictest-in_vals3-3.0-out_vals3] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_success[array_api_strictest-in_vals4-3.0-out_vals4] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_failure[array_api_strict] ERROR tests/main/test_elementwise.py::TestNanToNum::test_fill_value_failure[array_api_strictest] ERROR tests/main/test_elementwise.py::TestSinc::test_simple[array_api_strict] ERROR tests/main/test_elementwise.py::TestSinc::test_simple[array_api_strictest] ERROR tests/main/test_elementwise.py::TestSinc::test_dtype[array_api_strict-0] ERROR tests/main/test_elementwise.py::TestSinc::test_dtype[array_api_strict-(1+3j)] ERROR tests/main/test_elementwise.py::TestSinc::test_dtype[array_api_strictest-0] ERROR tests/main/test_elementwise.py::TestSinc::test_dtype[array_api_strictest-(1+3j)] ERROR tests/main/test_elementwise.py::TestSinc::test_3d[array_api_strict] - V... ERROR tests/main/test_elementwise.py::TestSinc::test_3d[array_api_strictest] ERROR tests/main/test_elementwise.py::TestSinc::test_device[array_api_strict] ERROR tests/main/test_elementwise.py::TestSinc::test_device[array_api_strictest] ERROR tests/main/test_elementwise.py::TestSinc::test_xp[array_api_strict] - V... ERROR tests/main/test_elementwise.py::TestSinc::test_xp[array_api_strictest] ERROR tests/main/test_elementwise.py::TestAngle::test_simple[array_api_strict] ERROR tests/main/test_elementwise.py::TestAngle::test_simple[array_api_strictest] ERROR tests/main/test_elementwise.py::TestAngle::test_basic[array_api_strict] ERROR tests/main/test_elementwise.py::TestAngle::test_basic[array_api_strictest] ERROR tests/main/test_elementwise.py::TestAngle::test_real[array_api_strict] ERROR tests/main/test_elementwise.py::TestAngle::test_real[array_api_strictest] ERROR tests/main/test_elementwise.py::TestAngle::test_complex[array_api_strict] ERROR tests/main/test_elementwise.py::TestAngle::test_complex[array_api_strictest] ERROR tests/main/test_elementwise.py::TestAngle::test_integral[array_api_strict] ERROR tests/main/test_elementwise.py::TestAngle::test_integral[array_api_strictest] ERROR tests/main/test_elementwise.py::TestAngle::test_2d[array_api_strict] - ... ERROR tests/main/test_elementwise.py::TestAngle::test_2d[array_api_strictest] ERROR tests/main/test_elementwise.py::TestAngle::test_device[array_api_strict] ERROR tests/main/test_elementwise.py::TestAngle::test_device[array_api_strictest] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_basic[array_api_strict] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_basic[array_api_strictest] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_integral[array_api_strict-int32] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_integral[array_api_strict-int64] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_integral[array_api_strictest-int32] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_integral[array_api_strictest-int64] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_complex[array_api_strict] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_complex[array_api_strictest] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_bool[array_api_strict] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_bool[array_api_strictest] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_device[array_api_strict] ERROR tests/main/test_elementwise.py::TestDeg2Rad::test_device[array_api_strictest] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_basic[array_api_strict] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_basic[array_api_strictest] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_integral[array_api_strict-int32] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_integral[array_api_strict-int64] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_integral[array_api_strictest-int32] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_integral[array_api_strictest-int64] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_complex[array_api_strict] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_complex[array_api_strictest] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_bool[array_api_strict] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_bool[array_api_strictest] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_device[array_api_strict] ERROR tests/main/test_elementwise.py::TestRad2Deg::test_device[array_api_strictest] ERROR tests/main/test_helpers.py::TestIn1D::test_no_invert_assume_unique[array_api_strict-fast path] ERROR tests/main/test_helpers.py::TestIn1D::test_no_invert_assume_unique[array_api_strict-slow path] ERROR tests/main/test_helpers.py::TestIn1D::test_device[array_api_strict] - V... ERROR tests/main/test_helpers.py::TestIn1D::test_xp[array_api_strict] - Value... ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-bool-True-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-int16-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-uint8-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-float32-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-float64-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-complex64-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-complex128-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-float32-1.0-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-float64-1.0-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-complex64-1.0-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-complex128-1.0-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-complex64-1j-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-complex128-1j-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-bool-1-False] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-int64-1.0-False] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strict-float64-1j-False] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-bool-True-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-int16-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-uint8-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-float32-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-float64-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-complex64-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-complex128-1-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-float32-1.0-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-float64-1.0-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-complex64-1.0-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-complex128-1.0-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-complex64-1j-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-complex128-1j-True] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-bool-1-False] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-int64-1.0-False] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_scalar[array_api_strictest-float64-1j-False] ERROR tests/main/test_helpers.py::TestAsArrays::test_scalar_vs_scalar[array_api_strict] ERROR tests/main/test_helpers.py::TestAsArrays::test_scalar_vs_scalar[array_api_strictest] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int8-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int16-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int32-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-int64-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint8-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint16-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint32-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-uint64-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float32-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-float64-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex64-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-complex128-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strict-bool-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int8-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int16-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int32-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-int64-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint8-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint16-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint32-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-uint64-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float32-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-float64-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex64-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-complex128-bool] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-int8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-int16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-int32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-int64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-uint8] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-uint16] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-uint32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-uint64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-float32] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-float64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-complex64] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-complex128] ERROR tests/main/test_helpers.py::TestAsArrays::test_array_vs_array[array_api_strictest-bool-bool] ERROR tests/main/test_helpers.py::test_eager_shape[array_api_strict] - ValueE... ERROR tests/main/test_helpers.py::TestMetaNamespace::test_basic[array_api_strict] ERROR tests/main/test_helpers.py::TestMetaNamespace::test_basic[array_api_strictest] ERROR tests/main/test_helpers.py::TestMetaNamespace::test_xp[array_api_strict] ERROR tests/main/test_helpers.py::TestMetaNamespace::test_xp[array_api_strictest] ERROR tests/main/test_helpers.py::TestCapabilities::test_basic[array_api_strict] ERROR tests/main/test_helpers.py::TestCapabilities::test_basic[array_api_strictest] ERROR tests/main/test_helpers.py::TestCapabilities::test_device[array_api_strict] ERROR tests/main/test_helpers.py::TestCapabilities::test_device[array_api_strictest] ERROR tests/main/test_indexing.py::TestDiagIndices::test_basic[array_api_strict] ERROR tests/main/test_indexing.py::TestDiagIndices::test_basic[array_api_strictest] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-1-2] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-1-4] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-1-7] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-2-2] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-2-4] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-2-7] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-3-2] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-3-4] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-3-7] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-4-2] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-4-4] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strict-4-7] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-1-2] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-1-4] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-1-7] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-2-2] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-2-4] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-2-7] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-3-2] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-3-4] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-3-7] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-4-2] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-4-4] ERROR tests/main/test_indexing.py::TestDiagIndices::test_ndim[array_api_strictest-4-7] ERROR tests/main/test_indexing.py::TestDiagIndices::test_empty[array_api_strict] ERROR tests/main/test_indexing.py::TestDiagIndices::test_empty[array_api_strictest] ERROR tests/main/test_indexing.py::TestDiagIndices::test_validation[array_api_strict] ERROR tests/main/test_indexing.py::TestDiagIndices::test_validation[array_api_strictest] ERROR tests/main/test_indexing.py::TestDiagIndices::test_device[array_api_strict] ERROR tests/main/test_indexing.py::TestDiagIndices::test_device[array_api_strictest] ERROR tests/main/test_indexing.py::TestTriIndices::test_basic[array_api_strict-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_basic[array_api_strict-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_basic[array_api_strictest-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_basic[array_api_strictest-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict--2-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict--2-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict--1-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict--1-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict-0-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict-0-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict-1-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict-1-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict-2-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strict-2-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest--2-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest--2-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest--1-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest--1-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest-0-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest-0-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest-1-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest-1-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest-2-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_offset[array_api_strictest-2-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_rectangular[array_api_strict-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_rectangular[array_api_strict-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_rectangular[array_api_strictest-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_rectangular[array_api_strictest-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_use_to_read[array_api_strict-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_use_to_read[array_api_strict-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_use_to_read[array_api_strictest-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_use_to_read[array_api_strictest-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_validation[array_api_strict-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_validation[array_api_strict-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_validation[array_api_strictest-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_validation[array_api_strictest-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_device[array_api_strict-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_device[array_api_strict-triu] ERROR tests/main/test_indexing.py::TestTriIndices::test_device[array_api_strictest-tril] ERROR tests/main/test_indexing.py::TestTriIndices::test_device[array_api_strictest-triu] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_simple[array_api_strict] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_simple[array_api_strictest] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_indices_scalar[array_api_strict] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_indices_scalar[array_api_strictest] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_indices_2d[array_api_strict] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_indices_2d[array_api_strictest] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_device[array_api_strict] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_device[array_api_strictest] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_xp[array_api_strict] ERROR tests/main/test_indexing.py::TestUnravelIndex::test_xp[array_api_strictest] ERROR tests/main/test_inspection.py::TestDefaultDType::test_basic[array_api_strict] ERROR tests/main/test_inspection.py::TestDefaultDType::test_basic[array_api_strictest] ERROR tests/main/test_inspection.py::TestDefaultDType::test_kind[array_api_strict] ERROR tests/main/test_inspection.py::TestDefaultDType::test_kind[array_api_strictest] ERROR tests/main/test_inspection.py::TestDefaultDType::test_device[array_api_strict] ERROR tests/main/test_inspection.py::TestDefaultDType::test_device[array_api_strictest] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strict-int32-shape0-False] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strict-int32-shape0-True] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strict-int32-shape1-False] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strict-int32-shape1-True] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strict-float64-shape0-False] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strict-float64-shape0-True] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strict-float64-shape1-False] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strict-float64-shape1-True] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strictest-int32-shape0-False] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strictest-int32-shape0-True] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strictest-int32-shape1-False] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strictest-int32-shape1-True] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strictest-float64-shape0-False] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strictest-float64-shape0-True] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strictest-float64-shape1-False] ERROR tests/main/test_lazy.py::test_lazy_apply_simple[array_api_strictest-float64-shape1-True] ERROR tests/main/test_lazy.py::test_lazy_apply_broadcast[array_api_strict-False] ERROR tests/main/test_lazy.py::test_lazy_apply_broadcast[array_api_strict-True] ERROR tests/main/test_lazy.py::test_lazy_apply_broadcast[array_api_strictest-False] ERROR tests/main/test_lazy.py::test_lazy_apply_broadcast[array_api_strictest-True] ERROR tests/main/test_lazy.py::test_lazy_apply_multi_output[array_api_strict-False] ERROR tests/main/test_lazy.py::test_lazy_apply_multi_output[array_api_strict-True] ERROR tests/main/test_lazy.py::test_lazy_apply_multi_output[array_api_strictest-False] ERROR tests/main/test_lazy.py::test_lazy_apply_multi_output[array_api_strictest-True] ERROR tests/main/test_lazy.py::test_lazy_apply_multi_output_broadcast_dtype[array_api_strict-False] ERROR tests/main/test_lazy.py::test_lazy_apply_multi_output_broadcast_dtype[array_api_strict-True] ERROR tests/main/test_lazy.py::test_lazy_apply_multi_output_broadcast_dtype[array_api_strictest-False] ERROR tests/main/test_lazy.py::test_lazy_apply_multi_output_broadcast_dtype[array_api_strictest-True] ERROR tests/main/test_lazy.py::test_lazy_apply_none_shape_in_args[array_api_strict] ERROR tests/main/test_lazy.py::test_lazy_apply_none_shape_in_args[array_api_strictest] ERROR tests/main/test_lazy.py::test_lazy_apply_none_shape_broadcast[array_api_strict] ERROR tests/main/test_lazy.py::test_lazy_apply_device[array_api_strict-False] ERROR tests/main/test_lazy.py::test_lazy_apply_device[array_api_strictest-False] ERROR tests/main/test_lazy.py::test_lazy_apply_device[array_api_strictest-True] ERROR tests/main/test_lazy.py::test_lazy_apply_arraylike[array_api_strict] - ... ERROR tests/main/test_lazy.py::test_lazy_apply_arraylike[array_api_strictest] ERROR tests/main/test_lazy.py::test_lazy_apply_scalars_and_nones[array_api_strict] ERROR tests/main/test_lazy.py::test_lazy_apply_scalars_and_nones[array_api_strictest] ERROR tests/main/test_lazy.py::test_lazy_apply_kwargs[array_api_strict-False] ERROR tests/main/test_lazy.py::test_lazy_apply_kwargs[array_api_strict-True] ERROR tests/main/test_lazy.py::test_lazy_apply_kwargs[array_api_strictest-False] ERROR tests/main/test_lazy.py::test_lazy_apply_kwargs[array_api_strictest-True] ERROR tests/main/test_lazy.py::test_lazy_apply_raises[array_api_strict] - Val... ERROR tests/main/test_lazy.py::test_lazy_apply_raises[array_api_strictest] - ... ERROR tests/main/test_linalg.py::TestKron::test_basic[array_api_strict] - Val... ERROR tests/main/test_linalg.py::TestKron::test_basic[array_api_strictest] - ... ERROR tests/main/test_linalg.py::TestKron::test_kron_smoke[array_api_strict] ERROR tests/main/test_linalg.py::TestKron::test_kron_smoke[array_api_strictest] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strict-shape_a0-shape_b0] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strict-shape_a1-shape_b1] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strict-shape_a2-shape_b2] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strict-shape_a3-shape_b3] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strict-shape_a4-shape_b4] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strict-shape_a5-shape_b5] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strictest-shape_a0-shape_b0] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strictest-shape_a1-shape_b1] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strictest-shape_a2-shape_b2] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strictest-shape_a3-shape_b3] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strictest-shape_a4-shape_b4] ERROR tests/main/test_linalg.py::TestKron::test_kron_shape[array_api_strictest-shape_a5-shape_b5] ERROR tests/main/test_linalg.py::TestKron::test_python_scalar[array_api_strict] ERROR tests/main/test_linalg.py::TestKron::test_python_scalar[array_api_strictest] ERROR tests/main/test_linalg.py::TestKron::test_device[array_api_strict] - Va... ERROR tests/main/test_linalg.py::TestKron::test_device[array_api_strictest] ERROR tests/main/test_linalg.py::TestKron::test_xp[array_api_strict] - ValueE... ERROR tests/main/test_linalg.py::TestKron::test_xp[array_api_strictest] - Val... ERROR tests/main/test_manipulation.py::TestAtLeastND::test_0D[array_api_strict] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_0D[array_api_strictest] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strict-input_shape0-0-expected_shape0] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strict-input_shape1-1-expected_shape1] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strict-input_shape2-2-expected_shape2] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strict-input_shape3-3-expected_shape3] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strict-input_shape4-5-expected_shape4] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strictest-input_shape0-0-expected_shape0] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strictest-input_shape1-1-expected_shape1] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strictest-input_shape2-2-expected_shape2] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strictest-input_shape3-3-expected_shape3] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_shapes[array_api_strictest-input_shape4-5-expected_shape4] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_values[array_api_strict] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_1D_values[array_api_strictest] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strict-input_shape0-0-expected_shape0] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strict-input_shape1-1-expected_shape1] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strict-input_shape2-2-expected_shape2] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strict-input_shape3-3-expected_shape3] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strict-input_shape4-5-expected_shape4] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strictest-input_shape0-0-expected_shape0] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strictest-input_shape1-1-expected_shape1] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strictest-input_shape2-2-expected_shape2] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strictest-input_shape3-3-expected_shape3] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_shapes[array_api_strictest-input_shape4-5-expected_shape4] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_values[array_api_strict] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_2D_values[array_api_strictest] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strict-input_shape0-0-expected_shape0] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strict-input_shape1-1-expected_shape1] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strict-input_shape2-2-expected_shape2] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strict-input_shape3-3-expected_shape3] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strict-input_shape4-5-expected_shape4] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strictest-input_shape0-0-expected_shape0] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strictest-input_shape1-1-expected_shape1] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strictest-input_shape2-2-expected_shape2] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strictest-input_shape3-3-expected_shape3] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_shapes[array_api_strictest-input_shape4-5-expected_shape4] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_values[array_api_strict] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_3D_values[array_api_strictest] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strict-input_shape0-0-expected_shape0] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strict-input_shape1-2-expected_shape1] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strict-input_shape2-5-expected_shape2] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strict-input_shape3-6-expected_shape3] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strict-input_shape4-9-expected_shape4] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strictest-input_shape0-0-expected_shape0] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strictest-input_shape1-2-expected_shape1] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strictest-input_shape2-5-expected_shape2] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strictest-input_shape3-6-expected_shape3] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_shapes[array_api_strictest-input_shape4-9-expected_shape4] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_values[array_api_strict] ERROR tests/main/test_manipulation.py::TestAtLeastND::test_5D_values[array_api_strictest] ERROR tests/main/test_manipulation.py::TestBroadcastShapes::test_xp[array_api_strict] ERROR tests/main/test_manipulation.py::TestBroadcastShapes::test_xp[array_api_strictest] ERROR tests/main/test_manipulation.py::TestExpandDims::test_single_axis[array_api_strict] ERROR tests/main/test_manipulation.py::TestExpandDims::test_single_axis[array_api_strictest] ERROR tests/main/test_manipulation.py::TestExpandDims::test_axis_tuple[array_api_strict] ERROR tests/main/test_manipulation.py::TestExpandDims::test_axis_tuple[array_api_strictest] ERROR tests/main/test_manipulation.py::TestExpandDims::test_axis_out_of_range[array_api_strict] ERROR tests/main/test_manipulation.py::TestExpandDims::test_axis_out_of_range[array_api_strictest] ERROR tests/main/test_manipulation.py::TestExpandDims::test_repeated_axis[array_api_strict] ERROR tests/main/test_manipulation.py::TestExpandDims::test_repeated_axis[array_api_strictest] ERROR tests/main/test_manipulation.py::TestExpandDims::test_positive_negative_repeated[array_api_strict] ERROR tests/main/test_manipulation.py::TestExpandDims::test_positive_negative_repeated[array_api_strictest] ERROR tests/main/test_manipulation.py::TestExpandDims::test_device[array_api_strict] ERROR tests/main/test_manipulation.py::TestExpandDims::test_device[array_api_strictest] ERROR tests/main/test_manipulation.py::TestExpandDims::test_xp[array_api_strict] ERROR tests/main/test_manipulation.py::TestExpandDims::test_xp[array_api_strictest] ERROR tests/main/test_manipulation.py::TestPad::test_simple[array_api_strict] ERROR tests/main/test_manipulation.py::TestPad::test_simple[array_api_strictest] ERROR tests/main/test_manipulation.py::TestPad::test_fill_value[array_api_strict] ERROR tests/main/test_manipulation.py::TestPad::test_fill_value[array_api_strictest] ERROR tests/main/test_manipulation.py::TestPad::test_ndim[array_api_strict] ERROR tests/main/test_manipulation.py::TestPad::test_ndim[array_api_strictest] ERROR tests/main/test_manipulation.py::TestPad::test_mode_not_implemented[array_api_strict] ERROR tests/main/test_manipulation.py::TestPad::test_mode_not_implemented[array_api_strictest] ERROR tests/main/test_manipulation.py::TestPad::test_device[array_api_strict] ERROR tests/main/test_manipulation.py::TestPad::test_device[array_api_strictest] ERROR tests/main/test_manipulation.py::TestPad::test_xp[array_api_strict] - V... ERROR tests/main/test_manipulation.py::TestPad::test_xp[array_api_strictest] ERROR tests/main/test_manipulation.py::TestPad::test_tuple_width[array_api_strict] ERROR tests/main/test_manipulation.py::TestPad::test_tuple_width[array_api_strictest] ERROR tests/main/test_manipulation.py::TestPad::test_sequence_of_tuples_width[array_api_strict] ERROR tests/main/test_manipulation.py::TestPad::test_sequence_of_tuples_width[array_api_strictest] ERROR tests/main/test_searching.py::TestSearchsorted::test_input_validation[array_api_strict] ERROR tests/main/test_searching.py::TestSearchsorted::test_input_validation[array_api_strictest] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-False-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-False-True-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-False-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted0-True-True-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-False-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-False-True-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-False-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strict-searchsorted1-True-True-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-False-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-False-True-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-False-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted0-True-True-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-False-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-False-True-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-False-shape10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-0-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-0-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-0-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-0-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-1-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-1-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-1-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-1-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-2-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-2-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-2-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-2-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-10-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-11-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-11-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-11-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-11-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-1000-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-1000-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-1000-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-1000-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-10001-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-10001-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-10001-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-10001-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape7-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape7-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape7-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape7-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape8-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape8-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape8-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape8-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape9-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape9-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape9-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape9-True-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape10-False-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape10-False-right] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape10-True-left] ERROR tests/main/test_searching.py::TestSearchsorted::test_nd[array_api_strictest-searchsorted1-True-True-shape10-True-right] ERROR tests/main/test_set.py::TestNUnique::test_simple[array_api_strict] - Va... ERROR tests/main/test_set.py::TestNUnique::test_simple[array_api_strictest] ERROR tests/main/test_set.py::TestNUnique::test_empty[array_api_strict] - Val... ERROR tests/main/test_set.py::TestNUnique::test_empty[array_api_strictest] - ... ERROR tests/main/test_set.py::TestNUnique::test_size1[array_api_strict] - Val... ERROR tests/main/test_set.py::TestNUnique::test_size1[array_api_strictest] - ... ERROR tests/main/test_set.py::TestNUnique::test_all_equal[array_api_strict] ERROR tests/main/test_set.py::TestNUnique::test_all_equal[array_api_strictest] ERROR tests/main/test_set.py::TestNUnique::test_nan[array_api_strict] - Value... ERROR tests/main/test_set.py::TestNUnique::test_nan[array_api_strictest] - Va... ERROR tests/main/test_set.py::TestNUnique::test_device[array_api_strict-0] - ... ERROR tests/main/test_set.py::TestNUnique::test_device[array_api_strict-1] - ... ERROR tests/main/test_set.py::TestNUnique::test_device[array_api_strict-2] - ... ERROR tests/main/test_set.py::TestNUnique::test_device[array_api_strictest-0] ERROR tests/main/test_set.py::TestNUnique::test_device[array_api_strictest-1] ERROR tests/main/test_set.py::TestNUnique::test_device[array_api_strictest-2] ERROR tests/main/test_set.py::TestNUnique::test_xp[array_api_strict] - ValueE... ERROR tests/main/test_set.py::TestNUnique::test_xp[array_api_strictest] - Val... ERROR tests/main/test_set.py::TestSetDiff1D::test_setdiff1d[array_api_strict] ERROR tests/main/test_set.py::TestSetDiff1D::test_assume_unique[array_api_strict] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape20-shape10-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape20-shape10-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape20-shape11-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape20-shape11-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape20-shape12-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape20-shape12-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape21-shape10-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape21-shape10-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape21-shape11-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape21-shape11-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape21-shape12-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape21-shape12-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape22-shape10-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape22-shape10-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape22-shape11-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape22-shape11-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape22-shape12-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_shapes[array_api_strict-shape22-shape12-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_python_scalar[array_api_strict-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_python_scalar[array_api_strict-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_device[array_api_strict-True] ERROR tests/main/test_set.py::TestSetDiff1D::test_device[array_api_strict-False] ERROR tests/main/test_set.py::TestSetDiff1D::test_xp[array_api_strict] - Valu... ERROR tests/main/test_set.py::TestIsIn::test_simple[array_api_strict] - Value... ERROR tests/main/test_set.py::TestIsIn::test_simple[array_api_strictest] - Va... ERROR tests/main/test_set.py::TestIsIn::test_device[array_api_strict] - Value... ERROR tests/main/test_set.py::TestIsIn::test_device[array_api_strictest] - Va... ERROR tests/main/test_set.py::TestIsIn::test_assume_unique_and_invert[array_api_strict] ERROR tests/main/test_set.py::TestIsIn::test_assume_unique_and_invert[array_api_strictest] ERROR tests/main/test_set.py::TestIsIn::test_kind[array_api_strict] - ValueEr... ERROR tests/main/test_set.py::TestIsIn::test_kind[array_api_strictest] - Valu... ERROR tests/main/test_set.py::TestUnion1d::test_simple[array_api_strict] - Va... ERROR tests/main/test_set.py::TestUnion1d::test_2d[array_api_strict] - ValueE... ERROR tests/main/test_set.py::TestUnion1d::test_3d[array_api_strict] - ValueE... ERROR tests/main/test_set.py::TestUnion1d::test_device[array_api_strict] - Va... ERROR tests/main/test_sorting.py::TestPartition::test_1d[array_api_strict] - ... ERROR tests/main/test_sorting.py::TestPartition::test_1d[array_api_strictest] ERROR tests/main/test_sorting.py::TestPartition::test_nd[array_api_strict-2] ERROR tests/main/test_sorting.py::TestPartition::test_nd[array_api_strict-3] ERROR tests/main/test_sorting.py::TestPartition::test_nd[array_api_strict-4] ERROR tests/main/test_sorting.py::TestPartition::test_nd[array_api_strictest-2] ERROR tests/main/test_sorting.py::TestPartition::test_nd[array_api_strictest-3] ERROR tests/main/test_sorting.py::TestPartition::test_nd[array_api_strictest-4] ERROR tests/main/test_sorting.py::TestPartition::test_input_validation[array_api_strict] ERROR tests/main/test_sorting.py::TestPartition::test_input_validation[array_api_strictest] ERROR tests/main/test_sorting.py::TestArgpartition::test_1d[array_api_strict] ERROR tests/main/test_sorting.py::TestArgpartition::test_1d[array_api_strictest] ERROR tests/main/test_sorting.py::TestArgpartition::test_nd[array_api_strict-2] ERROR tests/main/test_sorting.py::TestArgpartition::test_nd[array_api_strict-3] ERROR tests/main/test_sorting.py::TestArgpartition::test_nd[array_api_strict-4] ERROR tests/main/test_sorting.py::TestArgpartition::test_nd[array_api_strictest-2] ERROR tests/main/test_sorting.py::TestArgpartition::test_nd[array_api_strictest-3] ERROR tests/main/test_sorting.py::TestArgpartition::test_nd[array_api_strictest-4] ERROR tests/main/test_sorting.py::TestArgpartition::test_input_validation[array_api_strict] ERROR tests/main/test_sorting.py::TestArgpartition::test_input_validation[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_basic[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_basic[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_complex[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_complex[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_complex_with_weights[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_complex_with_weights[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_empty[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_empty[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_combination[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_combination[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_device[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_device[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_xp[array_api_strict] - Va... ERROR tests/main/test_statistical.py::TestCov::test_xp[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_batch[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_batch[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_bias[array_api_strict-True] ERROR tests/main/test_statistical.py::TestCov::test_bias[array_api_strict-False] ERROR tests/main/test_statistical.py::TestCov::test_bias[array_api_strict-0] ERROR tests/main/test_statistical.py::TestCov::test_bias[array_api_strict-1] ERROR tests/main/test_statistical.py::TestCov::test_bias[array_api_strictest-True] ERROR tests/main/test_statistical.py::TestCov::test_bias[array_api_strictest-False] ERROR tests/main/test_statistical.py::TestCov::test_bias[array_api_strictest-0] ERROR tests/main/test_statistical.py::TestCov::test_bias[array_api_strictest-1] ERROR tests/main/test_statistical.py::TestCov::test_bias_batch[array_api_strict-True] ERROR tests/main/test_statistical.py::TestCov::test_bias_batch[array_api_strict-False] ERROR tests/main/test_statistical.py::TestCov::test_bias_batch[array_api_strict-0] ERROR tests/main/test_statistical.py::TestCov::test_bias_batch[array_api_strict-1] ERROR tests/main/test_statistical.py::TestCov::test_bias_batch[array_api_strictest-True] ERROR tests/main/test_statistical.py::TestCov::test_bias_batch[array_api_strictest-False] ERROR tests/main/test_statistical.py::TestCov::test_bias_batch[array_api_strictest-0] ERROR tests/main/test_statistical.py::TestCov::test_bias_batch[array_api_strictest-1] ERROR tests/main/test_statistical.py::TestCov::test_correction[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_correction[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_correction_float[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_correction_float[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_axis[array_api_strict] - ... ERROR tests/main/test_statistical.py::TestCov::test_axis[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_frequency_weights[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_frequency_weights[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_weights[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_weights[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_both_weights[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_both_weights[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_batch_with_weights[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_batch_with_weights[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_axis_with_weights[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_axis_with_weights[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_axis_out_of_bounds[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_axis_out_of_bounds[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_weights_wrong_ndim[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_weights_wrong_ndim[array_api_strictest] ERROR tests/main/test_statistical.py::TestCov::test_weights_wrong_length[array_api_strict] ERROR tests/main/test_statistical.py::TestCov::test_weights_wrong_length[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMin::test_simple[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMin::test_simple[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMin::test_bigger[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMin::test_bigger[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMin::test_with_infinity[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMin::test_with_infinity[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMin::test_scalar[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMin::test_scalar[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMin::test_all_nan_slice_2d[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMin::test_all_nan_slice_2d[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMin::test_device[array_api_strict-None] ERROR tests/main/test_statistical.py::TestNanMin::test_device[array_api_strict-0] ERROR tests/main/test_statistical.py::TestNanMin::test_device[array_api_strict-1] ERROR tests/main/test_statistical.py::TestNanMin::test_device[array_api_strictest-None] ERROR tests/main/test_statistical.py::TestNanMin::test_device[array_api_strictest-0] ERROR tests/main/test_statistical.py::TestNanMin::test_device[array_api_strictest-1] ERROR tests/main/test_statistical.py::TestNanMin::test_xp[array_api_strict-0-expected_list0] ERROR tests/main/test_statistical.py::TestNanMin::test_xp[array_api_strict-1-expected_list1] ERROR tests/main/test_statistical.py::TestNanMin::test_xp[array_api_strictest-0-expected_list0] ERROR tests/main/test_statistical.py::TestNanMin::test_xp[array_api_strictest-1-expected_list1] ERROR tests/main/test_statistical.py::TestNanMax::test_simple[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMax::test_simple[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMax::test_bigger[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMax::test_bigger[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMax::test_with_infinity[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMax::test_with_infinity[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMax::test_scalar[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMax::test_scalar[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMax::test_all_nan_slice_2d[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMax::test_all_nan_slice_2d[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMax::test_device[array_api_strict-None] ERROR tests/main/test_statistical.py::TestNanMax::test_device[array_api_strict-0] ERROR tests/main/test_statistical.py::TestNanMax::test_device[array_api_strict-1] ERROR tests/main/test_statistical.py::TestNanMax::test_device[array_api_strictest-None] ERROR tests/main/test_statistical.py::TestNanMax::test_device[array_api_strictest-0] ERROR tests/main/test_statistical.py::TestNanMax::test_device[array_api_strictest-1] ERROR tests/main/test_statistical.py::TestNanMax::test_xp[array_api_strict-0-expected_list0] ERROR tests/main/test_statistical.py::TestNanMax::test_xp[array_api_strict-1-expected_list1] ERROR tests/main/test_statistical.py::TestNanMax::test_xp[array_api_strictest-0-expected_list0] ERROR tests/main/test_statistical.py::TestNanMax::test_xp[array_api_strictest-1-expected_list1] ERROR tests/main/test_statistical.py::TestNanSum::test_simple[array_api_strict] ERROR tests/main/test_statistical.py::TestNanSum::test_simple[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanSum::test_bigger[array_api_strict] ERROR tests/main/test_statistical.py::TestNanSum::test_bigger[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanSum::test_all_nan_slice[array_api_strict] ERROR tests/main/test_statistical.py::TestNanSum::test_all_nan_slice[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanSum::test_scalar[array_api_strict] ERROR tests/main/test_statistical.py::TestNanSum::test_scalar[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanSum::test_device[array_api_strict-None] ERROR tests/main/test_statistical.py::TestNanSum::test_device[array_api_strict-0] ERROR tests/main/test_statistical.py::TestNanSum::test_device[array_api_strict-1] ERROR tests/main/test_statistical.py::TestNanSum::test_device[array_api_strictest-None] ERROR tests/main/test_statistical.py::TestNanSum::test_device[array_api_strictest-0] ERROR tests/main/test_statistical.py::TestNanSum::test_device[array_api_strictest-1] ERROR tests/main/test_statistical.py::TestNanSum::test_xp[array_api_strict-0-expected_list0] ERROR tests/main/test_statistical.py::TestNanSum::test_xp[array_api_strict-1-expected_list1] ERROR tests/main/test_statistical.py::TestNanSum::test_xp[array_api_strictest-0-expected_list0] ERROR tests/main/test_statistical.py::TestNanSum::test_xp[array_api_strictest-1-expected_list1] ERROR tests/main/test_statistical.py::TestNanMean::test_simple[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMean::test_simple[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMean::test_bigger[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMean::test_bigger[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMean::test_all_nan_slice[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMean::test_all_nan_slice[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMean::test_scalar[array_api_strict] ERROR tests/main/test_statistical.py::TestNanMean::test_scalar[array_api_strictest] ERROR tests/main/test_statistical.py::TestNanMean::test_device[array_api_strict-None] ERROR tests/main/test_statistical.py::TestNanMean::test_device[array_api_strict-0] ERROR tests/main/test_statistical.py::TestNanMean::test_device[array_api_strict-1] ERROR tests/main/test_statistical.py::TestNanMean::test_device[array_api_strictest-None] ERROR tests/main/test_statistical.py::TestNanMean::test_device[array_api_strictest-0] ERROR tests/main/test_statistical.py::TestNanMean::test_device[array_api_strictest-1] ERROR tests/main/test_statistical.py::TestNanMean::test_xp[array_api_strict-0-expected_list0] ERROR tests/main/test_statistical.py::TestNanMean::test_xp[array_api_strict-1-expected_list1] ERROR tests/main/test_statistical.py::TestNanMean::test_xp[array_api_strictest-0-expected_list0] ERROR tests/main/test_statistical.py::TestNanMean::test_xp[array_api_strictest-1-expected_list1] ERROR tests/main/test_testing.py::TestAsNumPyArray::test_basic[array_api_strict] ERROR tests/main/test_testing.py::TestAsNumPyArray::test_basic[array_api_strictest] ERROR tests/main/test_testing.py::TestAsNumPyArray::test_device[array_api_strict] ERROR tests/main/test_testing.py::TestAsNumPyArray::test_device[array_api_strictest] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_equal_close_basic[array_api_strict-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_equal_close_basic[array_api_strict-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_equal_close_basic[array_api_strictest-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_equal_close_basic[array_api_strictest-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_shape_dtype[array_api_strict-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_shape_dtype[array_api_strict-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_shape_dtype[array_api_strict-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_shape_dtype[array_api_strict-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_shape_dtype[array_api_strictest-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_shape_dtype[array_api_strictest-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_shape_dtype[array_api_strictest-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_shape_dtype[array_api_strictest-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_namespace[array_api_strict-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_namespace[array_api_strict-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_namespace[array_api_strict-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_namespace[array_api_strict-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_namespace[array_api_strictest-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_namespace[array_api_strictest-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_namespace[array_api_strictest-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_namespace[array_api_strictest-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_shape[array_api_strict-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_shape[array_api_strict-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_shape[array_api_strict-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_shape[array_api_strict-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_shape[array_api_strictest-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_shape[array_api_strictest-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_shape[array_api_strictest-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_shape[array_api_strictest-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_dtype[array_api_strict-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_dtype[array_api_strict-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_dtype[array_api_strict-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_dtype[array_api_strict-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_dtype[array_api_strictest-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_dtype[array_api_strictest-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_dtype[array_api_strictest-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_dtype[array_api_strictest-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_scalar[array_api_strict-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_scalar[array_api_strict-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_scalar[array_api_strict-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_scalar[array_api_strict-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_scalar[array_api_strictest-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_scalar[array_api_strictest-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_scalar[array_api_strictest-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_check_scalar[array_api_strictest-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_close_tolerance[array_api_strict-int64] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_close_tolerance[array_api_strict-float64] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_close_tolerance[array_api_strictest-int64] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_close_tolerance[array_api_strictest-float64] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_less[array_api_strict] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_less[array_api_strictest] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_none_shape[array_api_strict-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_none_shape[array_api_strict-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_none_shape[array_api_strict-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_none_shape[array_api_strict-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_device[array_api_strict-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_device[array_api_strict-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_device[array_api_strict-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_device[array_api_strict-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_device[array_api_strictest-assert_equal] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_device[array_api_strictest-assert_close] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_device[array_api_strictest-assert_less] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_device[array_api_strictest-assert_close_nulp] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_close_nulp[array_api_strict] ERROR tests/main/test_testing.py::TestAssertEqualCloseLess::test_assert_close_nulp[array_api_strictest] ERROR tests/main/test_testing.py::test_lazy_xp_function[array_api_strict] - V... ERROR tests/main/test_testing.py::test_lazy_xp_function[array_api_strictest] ERROR tests/main/test_testing.py::test_lazy_xp_function_static_params[array_api_strict] ERROR tests/main/test_testing.py::test_lazy_xp_function_static_params[array_api_strictest] ERROR tests/main/test_testing.py::test_lazy_xp_function_cython_ufuncs[array_api_strict] ERROR tests/main/test_testing.py::test_lazy_xp_function_cython_ufuncs[array_api_strictest] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_lazy_xp_function_classes[array_api_strict] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_lazy_xp_function_classes[array_api_strictest] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_static_methods_preserved[array_api_strict] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_static_methods_preserved[array_api_strictest] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_static_methods_wrapped[array_api_strict] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_static_methods_wrapped[array_api_strictest] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_class_methods_wrapped[array_api_strict] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_class_methods_wrapped[array_api_strictest] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_circumvention[array_api_strict] ERROR tests/main/test_testing.py::TestLazyXpFunctionClasses::test_circumvention[array_api_strictest] ERROR tests/main/test_testing.py::test_lazy_xp_function_opaque_wrappers[array_api_strict] ERROR tests/main/test_testing.py::test_lazy_xp_function_opaque_wrappers[array_api_strictest] ERROR tests/main/test_testing.py::test_lazy_xp_function_recursive[array_api_strict] ERROR tests/main/test_testing.py::test_lazy_xp_function_recursive[array_api_strictest] ERROR tests/main/test_testing.py::test_lazy_xp_modules[array_api_strict] - Va... ERROR tests/main/test_testing.py::test_lazy_xp_modules[array_api_strictest] ERROR tests/main/test_testing.py::test_patch_lazy_xp_functions_iter[array_api_strict] ERROR tests/main/test_testing.py::test_patch_lazy_xp_functions_iter[array_api_strictest] ============== 2438 passed, 10549 skipped, 2296 errors in 27.46s =============== error: Bad exit status from /var/tmp/rpm-tmp.xQAiat (%check) Bad exit status from /var/tmp/rpm-tmp.xQAiat (%check) RPM build errors: Finish: rpmbuild python-array-api-extra-0.11.5-1.fc43.src.rpm Finish: build phase for python-array-api-extra-0.11.5-1.fc43.src.rpm INFO: chroot_scan: 1 files copied to /var/lib/copr-rpmbuild/results/chroot_scan INFO: /var/lib/mock/fedora-43-x86_64-1794037731.754822/root/var/log/dnf5.log INFO: chroot_scan: creating tarball /var/lib/copr-rpmbuild/results/chroot_scan.tar.gz /bin/tar: Removing leading `/' from member names ERROR: Exception(/var/lib/copr-rpmbuild/results/python-array-api-extra-0.11.5-1.fc43.src.rpm) Config(fedora-43-x86_64) 0 minutes 44 seconds INFO: Results and/or logs in: /var/lib/copr-rpmbuild/results INFO: Cleaning up build root ('cleanup_on_failure=True') Start: clean chroot INFO: unmounting tmpfs. Finish: clean chroot ERROR: Command failed: # /usr/bin/systemd-nspawn -q -M e296a95ec3bb4cc7ad8a7f6e3723e338 -D /var/lib/mock/fedora-43-x86_64-1794037731.754822/root -a -u mockbuild --capability=cap_ipc_lock --capability=cap_ipc_lock --bind=/dev/btrfs-control --bind=/dev/mapper/control --bind=/dev/fuse --bind=/dev/loop-control --bind=/dev/loop0 --bind=/dev/loop1 --bind=/dev/loop2 --bind=/dev/loop3 --bind=/dev/loop4 --bind=/dev/loop5 --bind=/dev/loop6 --bind=/dev/loop7 --bind=/dev/loop8 --bind=/dev/loop9 --bind=/dev/loop10 --bind=/dev/loop11 --resolv-conf=off --resolv-conf=off --console=pipe --setenv=TERM=vt100 --setenv=SHELL=/bin/bash --setenv=HOME=/builddir --setenv=HOSTNAME=mock --setenv=PATH=/usr/bin:/bin:/usr/sbin:/sbin '--setenv=PROMPT_COMMAND=printf "\033]0;\007"' '--setenv=PS1= \s-\v\$ ' --setenv=LANG=C.UTF-8 bash --login -c '/usr/bin/rpmbuild -ba --noprep --noclean --target x86_64 /builddir/build/originals/python-array-api-extra.spec' Copr build error: Build failed